Talos 0.20 – A self-hosted AI agent with a permission kernel

Talos released version 0.20.3, a stable update to its open-source, self-hosted AI agent for terminal and Telegram on macOS and Linux, including Raspberry Pi, that runs actions only after a deterministic permission kernel rules on them. The MIT-licensed agent supports local models via Ollama, Claude through the Claude Code CLI, or a configured API provider, and the 0.20.3 update adds native Telegram command blocks, migration from Hermes or OpenClaw into inactive staging with one validated imported skill activated at a time, and Talos Connections secret-free disabled connector drafts. The release retains the frozen operator contract, permissions and dependency locks unchanged, and the project states stable is not a claim of error-free operation or a substitute for reviewing permissions and keeping backups.

An autonomous agent you can hand a shell to, because it can prove what it will not do. | Terminal | Telegram | |---|---| <sub Terminal: real local run from the current source with Ollama. Telegram: simulated transport using text generated by the current Talos implementation; no private chat or token is included. Demo provenance https://github.com/talos-kernel/Talos/blob/main/assets/demo/README.md .</sub talos-agent.ch https://talos-agent.ch · Field manual https://talos-agent.ch/docs/ · FAQ https://talos-agent.ch/faq/ · Changelog https://github.com/talos-kernel/Talos/blob/main/CHANGELOG.md · Fork Talos Talos is an open-source, self-hosted AI agent for terminal and Telegram on macOS and Linux, including Raspberry Pi. Use Ollama for local models, Claude through the Claude Code CLI, or a configured API provider. Talos searches, writes and runs tools — but only after a deterministic permission kernel has ruled on the action. The model proposes. It never decides. Choose Allow once , Allow this task , or an exact-action standing approval; you do not have to approve every step individually. Sandboxed shell execution and an append-only audit trail keep actions bounded and reviewable. Talos is MIT-licensed; model-provider and hosting costs depend on your setup. The 0.20.3 stable release retains the frozen operator contract https://github.com/talos-kernel/Talos/blob/main/docs/beta-contract.md . Stable is not a claim of error-free operation or a substitute for reviewing permissions and keeping backups. Install Talos install · Run local models with Ollama https://talos-agent.ch/docs/ model · Set up Telegram https://talos-agent.ch/docs/ channels · Watch a recorded session https://talos-agent.ch/console.html · Install with Pinokio https://github.com/talos-kernel/talos-pinokio Apple Silicon on macOS 14+ with Homebrew 7+: install the CLI with Homebrew https://github.com/talos-kernel/Talos/blob/main/docs/homebrew.md . New to Talos? Try three small, checkable tasks https://github.com/talos-kernel/Talos/blob/main/docs/first-task.md : turn notes into a handover, save and read back a local note, and compare release notes without inventing missing facts. Already tried it? Share your first-run result https://github.com/talos-kernel/Talos/issues/new?template=first run.yml , including installation blockers. No star or positive review required. Latest signed release — 0.20.3 https://github.com/talos-kernel/Talos/releases/tag/v0.20.3 : this stable update shows concrete Telegram work: native command blocks for known diagnostic arguments, file and skill names, and separate measured result states. It retains safe agent migration, Talos Connections and labelled scheduled results. The operator contract, permissions and dependency locks remain unchanged. Emoji-only streaming prefixes wait for readable text, avoiding oversized checkmark flashes. Included in 0.20.3: migration from Hermes or OpenClaw can now fingerprint and apply an exact redacted plan into inactive staging, archive that staging for recovery, and activate one validated imported skill at a time. Talos Connections https://github.com/talos-kernel/Talos/blob/main/docs/connections.md adds secret-free disabled connector drafts and an inventory of the existing isolated-worker registry. Neither feature inherits accounts, credentials, permissions, schedules or plugin authority. See the migration contract https://github.com/talos-kernel/Talos/blob/main/docs/import-agents.md and release changelog https://github.com/talos-kernel/Talos/blob/main/CHANGELOG.md . Recent milestones include missions and projects https://github.com/talos-kernel/Talos/blob/main/docs/missions.md , managed skill creation and restore, hash-locked installer tooling, and Telegram receipts that say Reply done while separately counting still-running worker jobs. The current source also names safe work-trail details such as the skill or shell executable, and delivers scheduled results as one labelled notification without streaming their private work. See the full changelog https://github.com/talos-kernel/Talos/blob/main/CHANGELOG.md . Source versus release: main can contain newer changes. Signed installers and package managers follow their pinned release; older archives remain immutable. 0.20.3 includes the Beta.8 pip 26.2.2 hash lock and isolates pip from per-user configuration. When upgrading from Beta.7 or older with custom pip settings, use PIP CONFIG FILE=/dev/null talos update once; those older updaters still inherit pip configuration. - Run local models: connect Ollama on hardware you control, or use a configured cloud provider. - Work from terminal or Telegram: inspect projects, organise notes, prepare reports and run permitted tools. - Keep authority explicit: approve once, grant eligible steps for the current task, or maintain exact-action standing rules. Hard denials remain denials. - Inspect the evidence: review tool outcomes and the audit trail; do not mistake a confident answer for a successful action. Start with the model, cost and permission FAQ https://talos-agent.ch/faq/ . Talos is not Talos Linux or Cisco Talos. It is a self-hosted AI agent, with a stable 0.20.3 release. If this approach is useful, star the repository to bookmark it , try a small task and report a reproducible result https://github.com/talos-kernel/Talos/issues . Want to improve it? Fork Talos https://github.com/talos-kernel/Talos/fork and read the contribution guide https://github.com/talos-kernel/Talos/blob/main/CONTRIBUTING.md . Keep private configuration and credentials out of your fork. message ──▶ event log ──▶ reason ──▶ ╔══════════╗ ──▶ capability ──▶ execute ║ KERNEL ║ token │ ╚══════════╝ ▼ ▲ verify ──▶ receipt the only place effects are authorised | Authority | a token bound to exact arguments — valid once , for 30 seconds | | A tool without a target extractor | DENY by construction, not by a rule someone wrote | | Ships with | zero identities that may command it | | The shell | sandboxed, or it refuses to run at all | curl -fsSL https://talos-agent.ch/install.sh | less read it first curl -fsSL https://talos-agent.ch/install.sh | bash then run it The installer verifies the signature and the checksum, runs the full suite — and then stops . Nothing starts listening until you say so. Contents — in the order they matter Start here Why this exists why-this-exists · What it does not do what-it-does-not-do · Install install · Fork and build fork-and-build Using it A session in the terminal a-session-in-the-terminal · Work on the side work-on-the-side · What it remembers what-it-remembers · Seeing what it did seeing-what-it-did · Commands commands How it holds How the kernel decides how-the-kernel-decides · The autonomy dial the-autonomy-dial · Timed runs timed-runs · Announced plans announced-plans · The browser that only reads the-browser-that-only-reads · Delegating delegating · WhatsApp through your own broker whatsapp-through-your-own-broker · The MCP registry the-mcp-registry · Identity identity Evidence What a run looks like what-a-run-looks-like · Audit trail audit-trail · Tools tools · Architecture architecture · Roadmap roadmap Every capable agent eventually asks for shell access. At that moment you are trusting a language model with your machine, and the usual answers are unsatisfying: either it asks you to confirm everything and you stop reading the prompts by day three , or it runs free behind a regex blocklist and you hope. Talos takes a third position. Authority is a token, not a list. Every effect is authorised individually, bound to its exact arguments and targets, valid once, for thirty seconds. Forgetting to call the gate does not produce an unchecked effect — it produces no effect at all, because the raw runners are unreachable without a token. That design is testable, and it is tested: 277 adversarial scenarios run on every change and try to get an effect past the kernel. They are in redteam.py https://github.com/talos-kernel/Talos/blob/main/redteam.py . Read them before you trust anything written above. Stated plainly, because a security claim without its limits is marketing: - run shell needs a sandbox the platform can actually provide. It runs under bubblewrap on Linux and sandbox-exec on macOS. Where neither is available it refuses rather than running unprotected, so on such a host the shell tool is simply unavailable until the operator overrides it on purpose TALOS SANDBOX ALLOW UNCONFINED=1 . - It is not a multi-tenant security boundary. One operator, one machine. Anyone who can run code in the process can reach the token mint. - There is no gateway, no setup portal and no config.yaml . Each was considered and left out, for the same reason. A gateway terminates foreign identities and then claims channel:id to the kernel instead of proving it. Identity checking would then hang on a component outside the kernel. A web portal is a network service with its own authentication, so it is a second source of permission beside the allowlist. And a second config file would be two truths about who may command it, when the path floor protects exactly one. What weakens it is not the number of commands. It is a second way in. - It does not defend against a malicious model. It defends against a mistaken one, and against prompt injection arriving through tool output. Those are different threats. - Search needs no account. Without TALOS BRAVE API KEY it used to refuse; now it answers over DuckDuckGo the ddgs package, imported at call time . A key decides which provider answers, not whether one does. - It has three ways in, and all of them fetch Telegram long-polling, mail over IMAP, the WhatsApp broker queue over SSH . That is the rule, not an accident: an inbound webhook would need a port the world can reach, which turns "outbound only" into "publicly reachable". Mail sits at Trust.ASK — an address proves no account, so it may ask and receive answers but never approve anything. WhatsApp over the Cloud API stays delivery-only for the same reason; the broker channel below is the one that fetches. - It hears locally or not at all. hear transcribes a recording with faster-whisper on the machine it runs on — what was said is often the most private thing in a day, and sending it to somebody else's model to be understood is the one place where "runs on your machine" would be traded for convenience. It is an ordinary READ with the file as its target, so a recording under ~/.secrets/ is refused without the module knowing anything about secrets. The model loads on first use, not at boot. - It can see, on a real file. A photo you send is fetched into workspace/inbox/ and the note carries the path, so see image has something to point at — and the kernel has a target to judge. The fetch asks the allowlist first: the channel parses updates before the kernel has ruled on identity, so an unconditional fetch would let anyone who finds the bot write files to your disk. speak writes a WAV offline piper . - It can take one still out of a video. grab frame runs ffmpeg once and puts a single picture in the inbox; see image looks at it afterwards. Both the video and the picture are targets the kernel judges, so a video under ~/.secrets/ is refused exactly like a recording — without that, calling a file a video would have been the way around the secret floor. Where the picture lands is derived by the kernel, not chosen by the model, and it is deliberately one frame rather than a series: fanning ffmpeg out over a whole film and encoding every frame is how a small machine ends up unreachable. - It does not generate pictures or video, on purpose. Reading one is a READ with a target; making one is a paid call to somebody else's model, and it was removed rather than kept around waiting for credit. Seeing stays. - Fast evaluation is local and advisory. /eval <text uses a bounded deterministic classifier with no provider, key, network call or Vercel dependency. It reports a measured duration; it never grants, denies or changes a permission. The same local preflight runs once before every model-backed task, including background tasks and resumptions, and records only category, confidence and duration. It does not replace or shorten Claude, Gemini, Ollama or another configured provider call; the deterministic kernel remains the only authority. Requires Python 3.11+ on Linux or macOS and a model connection: your own API key, or an installed, signed-in Claude or Hermes CLI. The setup wizard offers the connections available on your machine. git clone https://github.com/talos-kernel/talos.git cd talos python3 -m venv .venv && . .venv/bin/activate pip install --require-hashes -r requirements.lock -r requirements-dev.lock export PATH="$PWD/bin:$PATH" makes talos available in this shell python -m talos setup terminal local identity + model; no Telegram needed python -m talos chat your first conversation Prefer Telegram? Run the bot wizard, then start the service: python -m talos setup token, identity and model python -m talos run the Telegram agent setup proves what it can rather than trusting what you type: the token goes to getMe , and your identity comes from a real message you send your own bot — not from a number you copy. A wrong token announces itself on the first poll; a wrong identity does not, and a stranger's identity never does. Setup preserves unrelated settings and writes each section atomically with mode 0600. It prints the configuration path and the next command. Use talos doctor for missing prerequisites and talos help <command for command help. A running service needs a restart after changing its file settings. Coming from Hermes or OpenClaw? talos import produces a read-only, redacted plan for portable profile files, memory documents and Agent Skills. Repeating its fingerprint with --apply copies that exact plan into private inactive staging; it neither overwrites active state nor inherits authority. Staging can be moved intact into a recovery archive, and an imported skill activates only by name through the existing managed-skill validator. Secrets are re-entered, permissions are approved again, and schedules/plugins/connections remain disabled review tasks. See the migration contract https://github.com/talos-kernel/Talos/blob/main/docs/import-agents.md . Talos Connections is the public connector surface built into Talos. It inventories the existing isolated-worker MCP registry and stores secret-free connector drafts in a disabled state. It bundles no account, endpoint, credential or private service, and a draft cannot run until the operator separately installs it and opts in through both allowlists. See the Connections contract https://github.com/talos-kernel/Talos/blob/main/docs/connections.md . Later, a single part can be redone without repeating the rest: python -m talos setup model switch what it thinks with python -m talos setup mail add the second way in IMAP python -m talos config list model find settings by name or description python -m talos config describe TALOS STATUS STYLE python -m talos config set TALOS STATUS STYLE expressive python -m talos config set TALOS MODEL claude-fable-5-1 python -m talos config set TALOS MODEL OVERRIDES '{"local-model": {"context window": 128000}}' a window or a price the catalogue does not know — never a provider, url or key; shown as your word python -m talos models --refresh force the same provider-catalogue sync now python -m talos import list supported agent-home previews python -m talos import hermes --source "$HOME/.hermes" --dry-run python -m talos import openclaw --source "$HOME/.openclaw" --dry-run python -m talos import hermes --source "$HOME/.hermes" --apply --confirm <fingerprint python -m talos connections list active worker registry + disabled drafts python -m talos status what it did last python -m talos chat a session here; approvals at a terminal python -m talos ask "how many …?" one turn from here, answer on stdout python -m talos events --tool run shell what happened — read-only, filterable python -m talos why 4831 why that was allowed or refused python -m talos review what this installation should change python -m talos report --out audit.txt what was done and what was refused With TALOS STATUS STYLE=expressive , Telegram keeps one live activity card with tool icons, elapsed time and an outcome for each step. It updates in place; the answer stays separate. Tool and plan JSON is hidden even after a prose introduction. Completed answers stay formatted after approval buttons, and Markdown tables become compact labelled rows. Approval prompts keep the exact command visible. /stop interrupts the run and /log opens its receipts. A finished conversation does not imply that a delegated background job has finished. Version 0.20.1 added phase-specific headings, code-styled tool identifiers and separate completed/failed counts to the expressive card. A quiet activity indicator uses the existing five-second heartbeat; TALOS STATUS ANIMATION=0 disables it. Approval waits stay still, and a long wait is labelled without inventing progress or an ETA. Version 0.20.2 added native shell code blocks for a small set of known diagnostic argument grammars. Unknown arguments, inline scripts and credentials are visibly omitted; service names, test paths and host aliases remain operator-facing metadata. These previews never grant permission. Approval prompts retain the exact action. Each card distinguishes In progress, Done, Failed and Not executed; missing results never become successful actions. Operator-defined OpenAI-compatible endpoints can be listed in the JSON file named by TALOS CUSTOM PROVIDERS default: data/custom-providers.json . Each entry needs name , base url and models ; optional env key names that endpoint's own key. These providers use the native API adapter, including runtime fallback. Without env key no Authorization header is sent; a named but missing key fails closed. Built-in provider names cannot be replaced. The isolated model worker does not accept agent-defined custom providers; it retains its own provider allowlist. The long-running service refreshes configured model catalogues in a background thread at startup and at most once per day; startup and replies never wait for it. A complete, validated OpenAI, Anthropic, Ollama or LM Studio listing may add new models and remove retired ones. Paginated, incomplete, CLI/OAuth or otherwise unproven listings are additive only. Authentication, network, HTTP, JSON and empty-list failures preserve the exact last-known-good catalogue. If a complete list retires the active model, Talos hides it from new selections and records the fact, but never switches a running conversation to another model silently. Empty API replies and stream errors stay failures throughout the configured fallback chain. If every permitted route fails, the task ends with an explicit error and keeps its context. A short promise to act gets one bounded follow-up; completed tool actions are never replayed. Failed channel polling backs off from 1 to 30 seconds and resets when the channel recovers, while messages received from healthy channels continue. talos health marks a channel degraded until a successful poll confirms recovery; error counts remain in the log. Startup and recovery each record one channel.ready transition, including empty successful polls, not an event for every idle poll. This is polling evidence, not proof that a process is still alive or every reply arrived. Cancellation follows the provider actually doing the work, including an active fallback. Model switching stays busy until that call finishes. Background tasks retain the same configured fallback routes with independent cancellation. After each successful tool receipt, a fresh short work announcement can receive one correction; repeating it without progress ends explicitly as unfinished. A full walkthrough — install, identity, the session, every command, and the ones that are missing on purpose — is at talos-agent.ch/docs https://talos-agent.ch/docs/ . ask is not a second way in. It is a channel like any other , with the same protocol and no special right. Two consequences follow, and both are the point of it. While the allowlist is empty the local caller is admitted as cli:<uid automatically, so the first run works. The moment the list names one identity it becomes exhaustive, and the uid must appear as cli:<uid exactly like a Telegram number. A shell next to the agent is not an argument for granting it anything. And the turn always runs under the unattended ceiling : a one-liner waits for nothing, so NEEDS HUMAN becomes DENY and says so. Approve in the chat, where somebody is actually looking. It also refuses to run inside the agent's own sandbox. Otherwise the agent could start talos ask from its shell and give itself orders — no channel, no foreign identity, nobody reading along. You do not need a fork to run or configure Talos. Keep a fork when you want to add a provider, polling channel, blueprint, gated tool, package or red-team case — work that is useful to your installation and may be useful upstream. 1. Fork Talos https://github.com/talos-kernel/Talos/fork on GitHub. 2. Clone your fork and keep this repository as upstream : git clone https://github.com/YOUR-ACCOUNT/Talos.git cd Talos git remote add upstream https://github.com/talos-kernel/Talos.git python3 -m venv .venv && . .venv/bin/activate pip install --require-hashes -r requirements.lock -r requirements-dev.lock 3. Pick a bounded first change from the good first issues https://github.com/talos-kernel/Talos/issues?q=is%3Aissue+is%3Aopen+label%3A%22good+first+issue%22 , create a branch and prove it locally: git switch -c feat/your-change python -m pytest tests/ -q python redteam.py 4. Open a pull request that says what changed, what you ran and — for anything near the kernel — what the change makes possible that was impossible before. The contribution guide https://github.com/talos-kernel/Talos/blob/main/CONTRIBUTING.md names the safe extension points and the higher bar for kernel changes. If you find a way past the kernel, do not demonstrate it in a public fork or issue; use the private process in SECURITY.md https://github.com/talos-kernel/Talos/blob/main/SECURITY.md . ask answers once and leaves. For anything that takes more than one turn there is talos chat — the same channel name, the same identity, the same conductor, the same kernel. The loop is line for line the one the service runs for Telegram; only the messages come from stdin instead of getUpdates . bash $ talos chat ◉ Talos / 0.20.3 Your terminal. Your rules. ────────────────────────────────────────────────────────────────────────── model claude-cli/claude-opus-4-1 policy autonomy 5 identity cli:1000 session approvals possible — you are at a terminal /help for commands, exit to leave you › what changed in the log today? The session line is the one to read: it says, before you ask for anything, whether an approval is even possible here — so a refusal later is explained in advance rather than guessed at. The unattended ceiling hangs on the terminal, not on the command. This was the contested decision. Applying ask 's rule to an interactive session would mean chat could never write anything; dropping it would mean talos chat < jobs.txt in a cron job looks exactly like a human. So it is measured rather than claimed: both stdin and stdout must be a real tty . A pipe, a redirect, a cron run has none, and there NEEDS HUMAN becomes DENY as before. Checking only stdin would let talos chat out.log pass as attended — nobody reads a question written to a file. The in-session commands are the same ones the messenger has, through the same command centre. A second vocabulary that only worked here is exactly the duplication this project has paid for elsewhere. Some questions take two minutes, and the conversation should not stand still for them. › /background go through /var/log and summarise today's errors Background 1 started: go through /var/log and summarise today's… id bg a1b2c3d4e5f6 · runs unattended, so anything needing approval is refused A background task is a scheduled run, not a helper carrying your rights. That is the whole security decision, and it avoids inventing a new concept. It is not a subagent: that one is born from model text and may therefore only read. Here a human typed the task, so its origin is the same as any other message. But nobody is sitting in front of it — so the ceiling from the timed run applies, and anything that would need approval is refused. Putting a question into a chat where a different conversation is running is the most reliable way to land a "yes" on the wrong thing. Its context is empty : the task and nothing else. Two runs sharing one history write into each other, and afterwards nobody can say which of them said what. The result does not flow back into the history either — it arrives as its own message, marked as a report. If no ceiling is wired, the task is refused rather than run uncapped. A forgotten parameter may only ever allow less. A running background task can be steered : delegate steer queues a course correction, and supersedes unfinished reasoning before its next action — the same seam where a typed correction enters a foreground run, only from its own mailbox. The instruction arrives as a framed turn "no additional rights" ; every tool call it provokes passes the same kernel under the same ceiling. Only the person and conversation that started the task may steer it, a delegated run may not steer at all, and anything without a step boundary in this process — a synchronous delegate , a worker job, a timed run — answers with a refusal rather than a pretend "ok". And everything stoppable stops at once: /stopall also /estop aborts the running thought, drains the queue, detaches every background task at its next step and discards its report, and discards every pending approval — it never approves one. Timed runs stay; a stop that deleted timers would create the next incident while ending this one. The reply is an honest balance per category, and a second /stopall says so. The conversation is kept per channel, in memory only, bounded by turns and by characters. When the bound is reached the middle is summarised and both ends stay verbatim — the head usually carries the actual task, the tail whatever "and that too" refers to. You: set up the backup for the NAS Agent: done — see /etc/backup.d Earlier summarised : discussed retention, settled on 30 days; the NFS mount needed the hard option You: and add the offsite copy The summary is labelled as one and never disguised as a verbatim turn — a summary that looks like something actually said is a claim about words nobody used, and the model could quote it back as if it were. It goes into the prompt as history, never into the standing instructions, and the transcript reaches the summariser explicitly framed as data : otherwise "summarise this" is the most convenient place for an injected line to become an instruction, and a permanent one, because the summary stays. The bound is not a comfort feature. If summarising fails, the oldest turns are dropped as before. A history that kept growing after a failed summary would turn a cost question into a leak — what was said weeks ago would go out again. The thin feeling with an agent like this does not come from missing verbs. It comes from not being able to see what it did and why. Both of these are read-only : python -m talos events --tool run shell what happened, filterable python -m talos why 4831 why that was allowed or refused why names the verdict, the rule that produced it, and the targets it applied to — and states that those targets were derived from the real arguments , never taken from what the model wrote. Then it shows the rest of the same run, because "refused, and then?" is the question that makes people stop reading logs. There is deliberately no talos undo beside them. /undo exists, and since talos chat it is reachable from the command line through the same path the messenger uses. A second one would be convenience against the design. config set refuses two whole classes of key , and that is the point of it. Secrets never go on a command line — that lands them in the shell history and in ps for every user of the machine. And TALOS ALLOWED PRINCIPALS , TALOS SECRETS ENV , TALOS MAIL AUTHSERV ID , every TALOS BASE URL <PROVIDER and TALOS WEB ALLOWED ADDRESSES are not settings but policy : whoever writes them does not have to talk the kernel into anything, they reconfigure it. Not settable, not even with a confirmation — a confirmation is exactly what gets clicked away. config get answers REDACTED for a secret whether it is set or not , so the command cannot be used to find out which doors a machine has. A key belongs to one provider. There is no single api key and no single base url; credentials.py keeps a route per provider — key and address in one piece, resolved at the moment of the call rather than frozen when the reasoner was built, because /model switches providers while the process keeps running. If the selected provider has no key, the switch is refused and the previous one stays active: no other provider's key stands in for it. Until 2026-08-05 one field held ANTHROPIC API KEY or OPENAI API KEY for every provider, so choosing openai-api sent an Anthropic key to OpenAI as a bearer token — a credential disclosure that looks like a typo, because the wrong recipient answers 401 . There is no default allowlist. TALOS ALLOWED PRINCIPALS must name your identity or the agent refuses to start. A shipped default would be a backdoor printed in the source. PolicyKernel.decide returns exactly three verdicts — ALLOW , NEEDS HUMAN , DENY — and the executor calls no runner before it has one. | Class | Examples | Verdict | |---|---|---| | Ordinary work | ls , git status , tests, builds, writes inside the workspace | runs | | Risky but recoverable | curl \| sh , git reset --hard , rm -rf <path | asks you | | Persistence & secrets | .bashrc , systemd units, ~/.ssh , the agent's own source | asks you | | Catastrophic | rm -rf / , mkfs , dd to a block device, fork bomb, reboot | refused | | Secret reads, system paths | cat /etc/passwd , anything under /etc /boot /usr /bin | refused | DENY returns before the approval check, so a "yes" cannot reach a hardline rule. A one-shot approval has a five-minute TTL, is bound to the exact request, and is re-verified against file hashes immediately before execution TOCTOU . When a human is asked, the approval surface offers four deliberately different grants: - Allow once approves one exact action. The approval is consumed when that action runs; a later request asks again. - Allow this task approves every approval-required action in the current foreground task until it ends or the operator stops it. It has no time limit, includes later commands and targets in that task, and does not carry into a new task. Kernel DENY still wins. - Always allow creates a standing rule for the exact action fingerprint — its command or kernel-derived target — not for the tool in general. It remains until the operator revokes it through /revoke . - Deny closes the request. Nothing runs, and a denial never becomes permission. The model can request an action, but it cannot choose a wider grant. The kernel still judges each action, even when the question itself is skipped by a standing rule. Targets are derived from real tool arguments, never from a field the model could omit. A tool without a target extractor is DENY by construction. /autonomy 0..5 sits above the kernel and can only tighten, never loosen . Level 5 is exactly the unfiltered kernel — not "anything goes". A dial that granted rights would be a second source of permission next to the kernel, which is the thing this design removes. | Level | Reads | Writes | Shell | |---|---|---|---| | 0 | refused | refused | refused | | 1 | asks | refused | refused | | 2 | free | refused | refused | | 3 | free | asks | asks | | 4 | free | free in workspace | asks | | 5 | free | kernel decides | kernel decides | The level survives restarts because it lives in the event log. An unreadable log drops to level 0, not to the last convenient value. /every 90 <task repeats on an interval; /every 0 8 MON-FRI <task takes a cron expression, because an interval can say "every 90 minutes" but never "weekdays at 08:00". An expression is a better clock, not an extra permission — what runs afterwards passes the same kernel. And it passes one ceiling more. During an unattended run NEEDS HUMAN becomes DENY : what may run without asking runs, everything else is reported rather than performed . Not parked until morning either — an approval question whose occasion is six hours old is how reflexive clicking starts. So a timed run may do strictly less than something you typed, which is the opposite of how cron usually works. A timed run stays quiet while it works. Its final result arrives once, labelled Scheduled run , with the job ID plus /schedules and /unschedule controls. This is separate from an interactive work trail: the detailed tool evidence remains available in the audit log without making a background run look like the current conversation. A blueprint can give a timed run a memory and a probe . With continuity: true the previous result is placed in front of the task as data — "yesterday 91 %, and today?" — and a run that ends in the same failure as its predecessor is logged instead of repeated into the chat. With monitor: true and a probe a shell command the probe is read before the model is asked: an unchanged output means no model call at all, with schedule.skipped unchanged in the log. The probe is an ordinary run shell of the task's principal through the same executor, kernel and sandbox, under the same unattended ceiling — no new permission. And a probe that fails, whatever the reason, fires the run: a broken sensor must not be the way to silence a watcher. For a task with several steps the agent may announce the sequence before it starts: ≡ 3 steps — collect the log, find the failing service, report The announcement changes nothing about permission. Every step still passes the kernel one at a time, exactly as if it had been asked for alone, and a step that needs approval still stops and asks. What the announcement does is bind the run : - the step budget shrinks from the house limit to what was announced — a three-step plan cannot become forty tool calls; - the first step that fails ends the run with a report of what ran, what stopped it, and what therefore did not happen — instead of the model improvising around the failure, which is how an agent turns a refusal into a bigger second attempt; - the plan is read once . A tool result is a stranger's text; if it could install a second, larger plan mid-run, prompt injection would be a way to buy budget. A step may carry a condition, and the condition is not evaluated by the model : PLAN: {"goal": "restart and confirm", "steps": {"intent": "restart the service", "check": "contains:rc=0"}, {"intent": "write the report", "check": "wrote:/tmp/report.md"} } This closes a real hole. run shell returns rc=1 and the executor still records DONE — because the tool ran. Whether the work succeeded lived only in the output text, and the only reader of that text was the model, which has an interest in the answer. A condition is checked in code, in the order announced, against the receipt of the step that ran: ok finished cleanly , contains:<text its output held that text , wrote:<path it wrote exactly there, as the kernel derived the target — not as the model claimed it . A condition can only ever withhold a confirmation. If one never comes true, the answer is delivered in full and the system adds its own line beneath it: ✕ 1/2 announced checks met — NOT confirmed done. Still open: step 2 it writes to /tmp/report.md . The honest limit: the model writes the conditions too. A met condition proves that the run's own stated expectation came true — not that the job was done well. Someone who sets themselves a trivial condition passes it. What the code guarantees is narrower and still worth a lot: that the condition was really evaluated rather than asserted, that it is shown to you word for word so you can judge its worth, and that the real count also lands in the event log — which the model cannot write to, and where a forged verdict line in its prose is contradicted. Conditions inspect only the receipt of their own step, never the state of the world. One that could read arbitrary files would be an oracle around the kernel "read /etc/shadow and tell me whether it contains root" , and an unknown condition is dropped rather than counted, so inventing vocabulary gains nothing. So planning here does the opposite of what it usually does: it makes a run more predictable and less powerful. That is deliberate, and it is the same inversion the unattended ceiling makes — a capability that arrives with more rights attached is not a capability this design accepts. Under that ceiling the two compose without knowing about each other: an unattended plan stops at the first step that would need a human, and says so. web fetch returns source; half the web is an empty shell without JavaScript. browse renders the page in real Chromium and hands back what a reader would see. What it deliberately cannot do is operate a page — no clicking, typing or submitting. A click has no derivable target: "click the third element" cannot be mapped to a resource you could bind a permission to, and a tool without a derivable target is DENY here by construction. That is the right answer rather than a shortfall — an agent that operates strangers' pages has a path to effect that no kernel catches afterwards. It is also fenced more tightly than a URL check alone can manage. Chromium is pinned to exactly the address guard url verified, and nothing else resolves at all: --host-resolver-rules="MAP ~NOTFOUND, MAP <host <ip " A redirect to another name goes nowhere, a script pulled from an ad network likewise, and a DNS rebind between check and fetch — the classic hole in every filter that inspects only the first name — has no effect. The cost is honest: pages that load their content from a CDN under a different name come back incomplete. Incomplete beats uncontrolled. Every call gets a throwaway profile: no cookies, no signed-in sessions, no history. A page Talos opens sees a factory-fresh browser. delegate hands one self-contained question to a second run — and that run may only read . No writing, no shell, nothing that needs approval. It uses the same executor, the same kernel and the same identity as its caller; what changes is a fourth ceiling that turns everything except reading into DENY for as long as it lasts. The reason is the same one that shapes timed runs. A delegated run is born from model text , not from someone typing. Giving it the rights of the run that started it would make delegation the most convenient way to produce an effect nobody asked for — a second source of permission wearing a different name. So delegation buys reach, never power: it is for looking things up without filling the main run's context with the search. What comes back is data, bounded like any tool result, never an instruction. The Cloud-API channel whatsapp.py can only deliver, because Meta's inbound path is a webhook — and a webhook is a port the world can reach. The broker channel wabroker.py gets inbound without breaking the rule: a listener on an operator-controlled WhatsApp broker reachable over SSH appends the messages routed to Talos to a JSONL queue, and Talos pulls that queue over SSH with a persisted byte cursor. Every connection goes outward, to a machine the operator controls, with the operator's own key. Nothing listens. Replies go back the same way — through the broker's send script, base64-safe; files travel as scp plus broker. The sender number comes out of the operator's own WhatsApp account rather than a text field, so the channel carries Trust.FULL , and the conversations live in the same whatsapp:<number namespace as before. Three env vars, all off by default: TALOS WA BROKER SSH the SSH target — empty means the channel does not exist , TALOS WA BROKER QUEUE and TALOS WA BROKER CLI DIR . Configured, the broker wins over the Cloud-API variant: both are named whatsapp , and this is the one that fetches. A failed poll is a loud channel.error , and the cursor never advances on failure — silence would look exactly like an empty queue. Operator details: docs/whatsapp-broker.md https://github.com/talos-kernel/Talos/blob/main/docs/whatsapp-broker.md . MCP is shipped — over the claude-worker seam, not into the agent. Talos itself never speaks MCP; the claude -p child inside a UID-separated worker job does, and Talos only writes the MCP config and passes it through. Which servers may exist at all is an operator decision, declared in data/mcp-servers.json gitignored, fail-closed like entities.json : version: 1 mandatory, absolute commands only, an env field costs the whole entry, sixteen servers at most. The socket frame carries server names , never commands or args. And it is gated twice: a server must appear in the registry file and in the worker's TALOS CLAUDE WORKER MCP SERVERS allowlist, or the request is refused by name before any job starts. There is no marketplace and no third-party code in the agent process — a curated starting point is examples/mcp-servers.json https://github.com/talos-kernel/Talos/blob/main/examples/mcp-servers.json , the full contract is docs/claude-worker.md https://github.com/talos-kernel/Talos/blob/main/docs/claude-worker.md . SOUL.md carries the agent's name and character . Its first heading is the name — change TALOS to ARGUS , restart, and the agent is renamed everywhere, including the header of the live display. There is no second place where the name lives. AGENTS.md carries durable operating discipline; USER.md carries stable preferences for the operator. All three reload on the next message, have explicit context limits, are protected by the persistence floor, and survive updates as operator-owned state. Missing AGENTS.md or USER.md is harmless; a missing or broken SOUL.md uses the safe fallback. A plain question gets a plain answer — no status chrome. The moment a tool runs, one message starts tracking it and stays afterwards as the receipt: ◉ Talos · 9s · step 2/100 ✓ shell command 2s ▸ write note — disk.md ◉ agent · ◈ thinking · ≡ plan announced · ▸ tool · ✓ done · ✕ failed · ⏸ waiting on you · ⛒ refused · ↩ rolled back. One glyph, one meaning, never in the prose of an answer. exec.intent is written before anything happens: exec.intent → approval.parked → approval.granted → exec.intent → exec.result A crash mid-action still leaves a trace. Every run carries the id of the token that permitted it. /log shows the last effects, /undo rolls back the last file change. Thirty-three, and every one of them passes the same gate. There is no privileged tool and no tool that skips the kernel — a tool without a target extractor is DENY by construction. | run shell | a command, sandboxed, or refused where no sandbox exists | | computer run / computer status | guarded computer use: bounded desktop actions and their status, each step through the gate | | remote exec | a command on another machine over ssh — operator-allowlisted hosts, always a human's yes, standing rules bind to exact host+command | | http request | any REST API — read methods through the SSRF-hardened door, state-changing methods always a human's yes, binding to exact method+URL | | git | clone / fetch / pull / push with credentials — every op a human's yes, binding to exact op+repo+remote; local git work stays in the sandboxed shell | | cli anything | a pinned CLI-Anything harness curated third-party CLI from the operator-owned registry — unknown harness or subcommand is DENY , the rest always a human's yes, binding to exact name+subcommand; registry format https://github.com/talos-kernel/Talos/blob/main/docs/cli-anything.md | | entity status | a known name resolved to an operator-configured fixed read-only probe | | mailbox list / mailbox read | the configured inbox, read-only through IMAP EXAMINE and BODY.PEEK; no sending, marking, moving or deletion | | read file / write file / undo last | ordinary work on the kernel-derived path, with atomic publication and a snapshot behind the write | | browse / web fetch / web search | render-only, guarded URL, keyless search by default | | see image / grab frame / hear / speak | a picture, one still out of a video, a recording, a voice | | read document | a PDF, Word, Excel or PowerPoint file turned into text — locally, framed as untrusted | | vault search / vault get / vault write note | a markdown knowledge base, if you point it at one | | session search | what was said in earlier turns | | delegate | a sub-run that can only read | | delegate code / delegate dag / delegate status | a bounded coding job — or a small acyclic graph of them — for a confined Claude worker: opt-in, off by default, writes only into a kernel-derived disposable workspace | | delegate agy | the same confined worker's second backend the agy CLI : same trust form and workspace, behind its own opt-in gates on agent and worker, no MCP | | delegate codex | bounded implementation and independent review with Codex: the same kernel and sandbox, isolated auth, verified stream completion and originating-chat notifications; setup https://github.com/talos-kernel/Talos/blob/main/docs/codex-worker.md | | delegate steer | a course correction into a running /background task — a framed turn that supersedes unfinished reasoning, never a new right | | agent consult | bounded advice from a second, operator-configured agent — data, never permission | | ask operator | the one way it can ask you something on purpose | | skill write | a new skill, written exactly once — and never without a human's yes | | skill view | inspect managed skills, supporting files, hashes and archived revisions | | skill manage | create, edit, patch, deactivate and restore skills; every change needs approval | Talos ships with no real entity names, hosts or service units. To enable entity-aware status checks, copy the neutral example and replace every placeholder with infrastructure you control: mkdir -p data cp examples/entities.json data/entities.json data/entities.json is runtime state: Git ignores it, updates preserve it, and Talos reads it as context rather than authority. entity status accepts only a configured entity name; the fixed URL or systemd user unit comes from this file and cannot be supplied by the model. Talos can turn verified successful workflows into reusable skills and improve them later. The prompt receives a compact catalog; skill bodies are read on demand. Skills remain untrusted guidance: every suggested action still passes the kernel. skill view lists the managed library, reads a skill or supporting file, and returns expected sha256 plus archived revisions. skill manage supports create , edit , patch , write file , remove file , delete and restore . Every mutation is declared irreversible: it needs operator approval, cannot receive attended auto-approval, and is denied during unattended runs. The original create-only skill write stays compatible. Changes require the current hash from skill view or "absent" for a new file . Patches require a unique match unless replace all is explicit. Prior content is archived under .history ; restoration verifies its hash. delete deactivates the skill by removing SKILL.md , retaining supporting files and history so it can be restored. It is not secure erasure. Text files may live under references/ , scripts/ , templates/ and assets/ ; scripts are stored as data and never run automatically. Files stay inside the managed root, private by default, and secret-shaped content and allowed-tools are refused. Concurrent manager calls are serialized. Hash checks reject stale edits, but they are not an atomic compare-and-swap against external editors: avoid editing the same file while approving a change. History is bounded to 200 revisions per skill; reaching the limit requires operator maintenance, never silent pruning. Two blueprints use it on a schedule. skill-distillation "every sunday 20:15" reads the week's event log and distils at most one repeated, proven workflow into a skill candidate — written through skill write when a human is there to approve, parked as a vault note under patterns/ when the unattended run is refused. And daily-reflection closes each day with a look at the same log — verified live against a real installation, schedules and ceiling included, not only in the suite. There is no image generation. That is a decision, not a gap: an agent that can produce photographs is a different conversation than one that can only look at them. The repository contains a generic QmpInputAdapter for bounded visual input over a pinned QMP transport. It is not tied to a guest distribution or desktop. A successful input receipt records dispatch only; it does not prove guest delivery or a visible result. The planned macOS deployment is deliberately narrower: one fixed x86 64 Debian guest with Hyprland , running under QEMU TCG on Apple silicon, not HVF. The intended boundary excludes inbound forwarding, bridged networking, host folders, clipboard sharing and host credentials. Shell, browser-protocol, file and routine operations remain outside this visual-only path. That deployment is not yet hardware-qualified. It requires a signed, operator-supplied VM bundle and a matching pre-provisioned disk; Talos does not bundle or download QEMU or a guest disk. Evidence collected for any earlier guest prototype is not reusable for this fixed contract. The ARM64 Linux/KVM Computer is a separate backend documented in the Computer guide https://github.com/talos-kernel/Talos/blob/main/docs/computer.md ; the planned Mac contract is described in the QMP VM boundary https://github.com/talos-kernel/Talos/blob/main/docs/qmp-debian-vm.md . In a session — the same set in the terminal and in the messenger, through one command centre. /stop /stopall /queue /status /new /retry /background · /pending /approve /deny /allowed /revoke · /log /undo /policy /autonomy /tools /whoami /version · /usage /model /eval /reasoning /debug On the command line — each answering a question an operator actually asks: | chat | a session here; approvals at a terminal | | ask "…" | one turn, for scripts and cron | | setup · doctor · config · models | set it up, and find out why something won't work | | import list · import <hermes\|openclaw … | preview, stage, activate one skill, or archive rollback | | connections list · connections draft … | isolated-worker inventory and disabled secret-free connector drafts | | status · events · why <id | what it did, and why that was allowed or refused | | report · review | a record someone else can read · what this installation should change | | update · version | signature checked, tests first | /policy <path|command is a dry run: it calls decide and shows the verdict without executing anything. It is the fastest way to understand the kernel. /eval <text is a separate local diagnostic. The local classifier reports a likely category and explicitly does not establish truth. Its result is never consulted by decide and cannot approve an action. Small modules on purpose. The gate path policy.py , 996 lines has to be readable in one sitting — a gate you cannot read is not a gate. | Module | Role | |---|---| | policy.py | the kernel: target derivation, three verdicts | | capability.py | tokens: mint, redeem, invalidate | | command floor.py | hardline and dangerous command detection | | approval.py / standing.py | one-shot and standing approvals | | autonomy.py / trust.py / channel.py | ceilings above the kernel | | executor.py / verifier.py / snapshot.py | write-ahead execution, TOCTOU, undo | | conductor.py / agent loop.py / worker.py | orchestration | | plan.py / schedule.py / subagent.py | announced sequences, timed runs, delegation — all only tighten | | web.py / browser.py | the only door out: URL guard, then a resolver cage | | reasoner.py / provider.py | pluggable model backends | | intelligence.py / evaluation.py | bounded entity context, working state, factual review, adaptive reasoning and trace regression metrics | | credentials.py | one route per provider — key and address, resolved at call time | | sandbox.py | bubblewrap / sandbox-exec ; refuses rather than running unconfined | | identity.py / ux.py | name and glyphs | | telegram.py / mail.py / whatsapp.py / wabroker.py | channels — every way in fetches, none listens | | vision.py / hearing.py / speech.py / frames.py | reading a picture, hearing a recording, speaking, one still out of a video — ordinary READ/WRITE with a target | | cli.py / doctor.py / configcli.py / schema.py | the subcommands: diagnose, read and change settings — the schema decides what may be written, and the allowlist and the network exceptions never are | | askcli.py | talos ask — one turn from a script, as a channel with no special right | | import agents.py / import apply.py | bounded migration plans, inactive atomic staging, recovery archive and managed skill activation | | connections.py / mcpservers.py | public disabled connection drafts and the operator-owned isolated-worker registry | | models.py / model catalog sync.py | provider-bound live model lists; complete snapshots may replace membership, incomplete ones only add, and failures keep the last good cache | | updater.py | update beside the old tree, both suites in the new one, switch only if green | | eventlog.py / memory.py / usage.py | durable log, conversation memory, metering | | skillwrite.py | the one gated write into the skills directory | The beta release evidence https://github.com/talos-kernel/Talos/blob/main/docs/beta-readiness.md separates source tests, signed artifact checks and the seven-day operating observation. The frozen beta contract https://github.com/talos-kernel/Talos/blob/main/docs/beta-contract.md , recovery runbook https://github.com/talos-kernel/Talos/blob/main/docs/recovery.md and local python scripts/beta-gate.py --previous <published-version make these checks repeatable. The stable qualification record and gates https://github.com/talos-kernel/Talos/blob/main/docs/stable-readiness.md cover the transition from RC.3 to 0.20.0 , including seven days of deployed observation and separate final-artifact checks. The path to 1.0 remains a separate compatibility commitment supported by operating and upgrade evidence; it is not a calendar promise. 1. openat2 with RESOLVE BENEATH instead of realpath checks. 2. An adapter seam for tools. Adding a channel is already just implementing a protocol; adding a tool still touches three places, one of them inside the kernel TARGET EXTRACTORS . The seam should make a new effect routine without making the kernel optional — everything still passes the same gate, the same tokens, the same snapshots, the same receipts. 3. Ownership as the boundary, not just the floor. While the agent and its config file belong to the same writable user, the floor and the sandbox are code in the same process. A separate user for the model worker would make it a real separation. On a hardened installation the config already belongs to root and the agent may only read it. Landed since this list was first written: the sandbox for run shell bubblewrap / sandbox-exec , see What it does not do what-it-does-not-do , streaming replies, timed runs under a ceiling that makes them weaker than typed ones, announced plans announced-plans built so that planning tightens a run instead of widening it, read-only delegation, a render-only browser, the media tools, hearing on the machine itself, one still out of a video, keyless search, mail as a second way in , a command line that grants nothing by being one, MCP over the claude-worker seam an operator-owned registry the-mcp-registry instead of a marketplace , WhatsApp through an operator-controlled broker whatsapp-through-your-own-broker , and skills the agent may write — but never alone skills-it-writes-for-itself . Start with Fork and build fork-and-build and the good first issues https://github.com/talos-kernel/Talos/issues?q=is%3Aissue+is%3Aopen+label%3A%22good+first+issue%22 . Provider adapters, blueprints, packaging, tests and documentation are the most useful first contributions because they improve the agent without making the gate harder to audit. Changes to the kernel policy.py , capability.py , command floor.py , approval.py , standing.py , autonomy.py , trust.py , verifier.py , executor.py need: 1. pytest , redteam.py and e2e.py green, in that order. 2. Any loosening must add a red-team case proving the boundary next to it still holds. If you find a way past the kernel, that is the most valuable thing you can contribute — and it belongs in the private channel, not in a public issue: see SECURITY.md https://github.com/talos-kernel/Talos/blob/main/SECURITY.md , which also states the scope and the limits no patch will remove. Everything else — a missing tool, a refusal you disagree with, a wrong answer — is a normal issue and welcome as one. MIT. See LICENSE https://github.com/talos-kernel/Talos/blob/main/LICENSE . Telegram receipts distinguish a finished reply from ongoing delegated work: Reply done can include a count of pending background jobs whose results arrive separately.