Zerostack – An agent that's less sloppy than Elon's attempts

Minimal coding agent written in Rust, inspired by pi and opencode.

blogposts: what we built in 2 weeks memory design subagents design xavier's memory analysis

note: Want to support? Consider donating here; if you are a company interested in sponsoring zerostack, contact me here.

Multi-provider: OpenRouter, OpenAI, Anthropic, Gemini, Ollama, plus custom providers** Standard tools**: all of the standard tools exposed to coding agents, as described by the opencode documentation.** Permission system**: five configurable modes with per-tool patterns, session allowlists, and configurable mode-to-rule application policies** Session management**: save/load/resume sessions, auto-compaction to stay within context windows** Terminal UI**: crossterm-based, markdown rendering, mouse selection/copy, scrollback, reasoning visibility toggle** Prompts system**: switch between system prompt modes at runtime (code

,plan

,review

,debug

, etc.) to tailor the agent's behavior to the task without having to manage Skills.MCP support: connect MCP servers for extended tooling (exposed as an optional compile-time feature)** Integrated Exa search**: allows for WebFetch and WebSearch tools** Integrated Ralph Wiggum loops**: looping capabilities for long-horizon tasks** Integrated Git Worktrees integration**: Use/worktree

to move the agent from one worktree to another.ACP support(gated): Agent Communication Protocol server — lets editors (Zed, etc.) connect to zerostack as an ACP agent** Persistent memory**(gated): plain-Markdown memory across sessions: a global MEMORY.md plus per-project daily logs, scratchpad, and notes, injected into the system prompt each sessionLifecycle hooks(gated): observe or gate tool calls, prompts, and session lifecycle events via external commands, using asettings.json

schema largely compatible with Claude Code hooksAdvisor(gated): a second model the agent can consult mid-session for strategic guidance, with an optional human-handoff mode** Multimodal input**(gated): attach images and PDFs to messages** Subagents**: Parallel and fast, used for exploring the codebase** ARCHITECTURE.md**: Our own companion file for AGENTS.md, it allows to offer a shared core knowledge for all agents working on the same codebasePrompt chaining: offers to advance brainstorm → plan → code → review as each phase finishes, config-gated per transition** Status signals**: emits start/stop/git-conflict events over a Unix socket for external status bars or tooling

NOTE: Windows support is not tested is any way, but feel free to try and open an issue if you encounter any bugs!

zerostack is one of the smallest and most performant coding agents on the market.

• Lines of code: ~30k LoC (core, excluding tests)
• Binary size: 26MB
• RAM footprint: ~16MB on average, with peaks at ~24MB (vs ~300MB with peaks at ~700MB for opencode or other JS-based coding agents)
• CPU usage: 0.0% on idle, ~1.5% when using tools (measured on an Intel i5 7th gen, vs ~2% on idle and ~20% when working for opencode)

You can now read a complete Get Started guide for zerostack here!

curl -fsSL https://raw.githubusercontent.com/gi-dellav/zerostack/main/install.sh | bash

Or pick a tarball manually from GitHub Releases.

cargo install zerostack

cargo install zerostack --all-features

cargo install zerostack --features acp,memory,hooks,advisor
brew tap gi-dellav/tap
brew trust gi-dellav/tap   # required for Homebrew 6.0.0+
brew install zerostack

Run directly with nix-run:

$ nix-run https://github.com/gi-dellav/zerostack/archive/refs/heads/main.tar.gz

Add to profile:

$ nix profile add --file https://github.com/gi-dellav/zerostack/archive/refs/heads/main.tar.gz

Add as an overlay to your system/project:

let
  pkgs = import nixpkgs {
    overlays = [
      (import "${zerostack-src}/nix/overlay")
    ];
  };
in
pkgs.zerostack

Once installed, run /prompt autoconfig

inside zerostack to explore the documentation and configure the tool interactively.

note: If you have questions or you want to collaborate on the project, please join the dedicated Matrix chatroom.

If you want to orchestrate multiple zerostack agents from the terminal, also install multistack.

Install bubblewrap for --sandbox

, which runs every bash command inside an isolated environment to protect your system from accidental or malicious damage:

apt install bubblewrap

dnf install bubblewrap

pacman -S bubblewrap

There is also support for zerobox as an alternative sandbox backend.

export OPENROUTER_API_KEY="[api_key]"

zerostack

zerostack --no-color

zerostack -p "Explain this project"

zerostack -c

zerostack --provider openrouter --model deepseek/deepseek-v4-flash

See docs/CONFIG.md for config file location, accepted keys, provider aliases, permission rules, and MCP server configuration.

You can run /prompt autoconfig

in order to use a specialized agent that allows to navigate the documentation and customize your zerostack setup.

zerostack includes a set of built-in system prompts that change the agent's behavior and tone. The idea is to build a complete suite of prompts that can fully substitute skills like superpower or the Claude's official skills. You can switch between different prompts or list all registered prompts using /prompt

.

Built-in prompts:

Prompt	Description
(default)code
Coding mode with full file and bash tool access, TDD workflow
plan
Planning-only mode — explores and produces a plan without writing code
review
Code review mode — reviews for correctness, design, testing, and impact
debug
Debug mode — finds root cause before proposing fixes
ask
Read-only mode — only read/grep/find_files permitted, no writes or bash
brainstorm
Design-only mode — explores ideas and presents designs without code
frontend-design
Frontend design mode — distinctive, production-grade UI
review-security
Security review mode — finds exploitable vulnerabilities
simplify
Code simplification mode — refines for clarity without changing behavior
write-prompt
Prompt writing mode — creates and optimizes agent prompts
refactor
Refactoring mode — restructures code for design and maintainability while preserving behavior
autoconfig
Configuration mode — reads docs and edits your config/prompts, writes no code
orchestrator
Orchestration mode — combines direct tool use with parallel zerostack subprocess invocations for heavier work
write-text
Prose-writing mode — drafts and reviews non-code writing (docs, posts, emails)

You can also create custom prompts by placing markdown files in $XDG_CONFIG_HOME/zerostack/prompts/

and referencing them by name.

Additionally, the agent automatically loads AGENTS.md

or CLAUDE.md

from the project root or any ancestor directory, injecting their contents into the system prompt. When enabled (feature archmd

), ARCHITECTURE.md

is also loaded the same way, providing high-level design context to speed up exploration. Use -n

/ --no-context-files

to disable all context file .

When brainstorm

, plan

, or code

finishes, zerostack can offer to advance to the next phase (brainstorm

→ plan

→ code

→ /review

), asking Continue to plan? [Y/N/B]

-style at each step: Y

advances, N

stays put, and B

("but ...") advances with an extra instruction appended. Each transition is enabled independently in config.

zerostack has five permission modes:

Mode	CLI flag	Behavior
restrictive
-R / --restrictive
Ask for every operation. Config rules are ignored by default (can be enabled via permission-modes ).
readonly
--read-only
Allow read/grep/find_files/list_dir. Deny writes, edits, bash, and everything else. Config rules ignored by default.
guarded
--guarded
Allow read tools. Ask for writes, edits, bash, and everything else. Config rules apply.
standard
(default)	Allow path tools (read/write/edit/list_dir) within CWD and subdirectories. Safe bash commands (ls, cat, git log, cargo check) auto-allowed. Ask for external paths and unrecognized commands. Config rules apply and override mode defaults.
yolo
--yolo
Allow everything, but prompt for destructive bash commands (rm, dd, mkfs, etc.). Config rules apply.

The --dangerously-skip-permissions

flag completely bypasses all permission checks, allowing every tool operation without any guard. This is not a mode and cannot be toggled at runtime.

Permissions can be configured per-tool with granular glob patterns in the config file. For example, you can allow write **.rs

automatically while always asking before writing to other files.

A session allowlist persists approved decisions for the duration of the session, so you don't have to repeatedly confirm the same operation.

Doom-loop detection: identical tool calls repeated 3+ times trigger a warning prompt (or denial depending on your config), preventing runaway agents from spamming destructive operations.

This is a list of the most important slash commands:

/model

— Switch model/thinking

— Set thinking level/clear

— Clear conversation/session

— List/save/load sessions/loop

— Schedule recurring prompts/prompt

— List or change the agent's prompt/mode

— Set the permission system's mode/queue

— Manage input queued while the agent is busy/btw

— Ask a quick side question in parallel without interrupting the agent/review

— Run a one-shot code review in readonly mode, then restore the previous prompt/hooks

(gated) — Show whether a hook dispatcher is installed and what it's configured for/advisor

(gated) — Show or change advisor status (enabled, mode, model, max uses)

To see all of the commands, use /help

.

You can keep typing while the agent is running. Plain text is not sent right away and never starts a second concurrent run; it is queued and replayed as the next prompt once the current run finishes. Each queued line is shown as queued: <text>

.

Manage the queue with /queue

, which works even while a run is active:

/queue ls

lists the pending inputs (bare/queue

does the same)/queue clear

empties the queue/queue pop

removes the last queued input, to undo a mis-typed line

Selecting /queue

in the command picker opens a second-level menu with these three subcommands, so you do not need to remember them.

Commands (input starting with /

, .

, or !

) are not queued while a run is active: wait for it to finish, or press Ctrl-C. Ctrl-C cancels the running agent for real, including any child processes it spawned, and clears the queue.

/btw <message>

asks a quick "by the way" question in parallel with the main agent, without interrupting it. Like /queue

, it works even while the agent is busy. It forks the current context (including a trace of the agent's in-flight turn, when one is running) and answers using four read-only tools (read

, grep

, find_files

, list_dir

); it cannot write files or run commands. It then prints the reply inline. Nothing is written to conversation history, and its token usage is tracked separately in the status bar as btw:…

. Press Ctrl-C to cancel an in-flight /btw

without disturbing the main agent.

You can point a question at a specific file with @

: pick /btw

from the command menu, then type @

to open the file picker (for example /btw

then @src/main.rs

then "how does this work?"), and /btw

reads the file you reference.

Sessions are saved to $XDG_DATA_HOME/zerostack/sessions/

. Use -c

to resume the most recent session, -r

to browse and select one, or --session <id>

to load a specific session.

NOTE: Memory is gated behind the memory

feature and is not included in the default build. Install with cargo install zerostack --features memory

.

With the memory

feature, zerostack keeps plain-Markdown notes on disk and injects the relevant ones into the system prompt at the start of every session, so it remembers your preferences and recent context across runs.

Global memory files are stored in $XDG_DATA_HOME/zerostack/agent/memory/

.

NOTE: Hooks are gated behind the hooks

feature and are not included in the default build. Install with cargo install zerostack --features hooks

.

With the hooks

feature, external commands can observe or gate agent behavior at defined lifecycle events: a tool call (PreToolUse

/PostToolUse

), a user prompt (UserPromptSubmit

), the agent finishing a turn (Stop

), a session starting/ending, or a subagent starting/stopping. Hooks use the same settings.json

shape, stdin envelope, and exit-code/stdout-JSON contract as Claude Code, so an existing CC hooks setup is largely compatible.

Hook config lives in settings.json

at up to three locations (global, project, and an admin-managed file), merged in that order. The project-level file is untrusted by default; zerostack asks for confirmation (tracked by hash) before running project hooks for the first time. Use --no-hooks

to disable all non-managed hooks, or --hooks-test <tool>

to dry-run PreToolUse

hooks for a tool without starting a session. See docs/CONFIG.md for the full configuration reference.

NOTE: Advisor is gated behind the advisor

feature and is not included in the default build. Install with cargo install zerostack --features advisor

.

With the advisor

feature, the agent can consult a second model mid-session for strategic guidance (architecture calls, edge cases, course correction) without derailing the main coding session. Enable it with --advisor

, or toggle it at runtime with /advisor on

/ /advisor off

. An optional human-handoff mode (/advisor handoff on

) routes advisor calls to you instead of a model. See docs/CONFIG.md for configuration.

NOTE: Multimodal input is gated behind the multimodal

feature (images) and pdf

feature (PDFs, implies multimodal

), neither included in the default build.

With these features enabled, you can attach images and PDFs to your messages (up to 20 MB per file) and the agent processes them via the underlying provider's multimodal support.

NOTE: Status signals require the status-signals

feature, which is included in the default build.

Pass --status-socket <path>

to have zerostack emit start

, stop

, and git-conflict

events over a Unix domain socket, for external status bars or tooling to watch.

If you want to make multiple agents work on the same repository without having to work with git worktrees, zerostack now ships with --parallel

, which enables full management of a temporary git worktree that will be merged and removed before exiting the agent.

zerostack includes an iterative coding loop for long-horizon tasks. The agent repeatedly reads the task, picks an item from the plan, works on it, runs tests, updates the plan, and loops until the task is complete or the iteration limit is reached.

NOTE The loop system is an experimental feature.

/loop Implement the user authentication system
/loop stop
/loop status

/loop <prompt>

— Start a loop with the given prompt/loop stop

— Stop the active loop/loop status

— Show current loop state

Each iteration includes the original task, the evolving LOOP_PLAN.md

, a summary of the previous iteration, and any validation output. Non-slash input is blocked while a loop is active.

zerostack --loop --loop-prompt "Refactor the API" --loop-max 10 --loop-run "cargo test"

Flag	Description
--loop
Enable headless loop mode
--loop-prompt <text>
Prompt for each iteration
--loop-plan <path>
Custom plan file path (default: LOOP_PLAN.md )
--loop-max <N>
Maximum iterations (default: unlimited)
--loop-run <cmd>
Validation command to run after each iteration

zerostack provides a branch-per-task workflow using git worktrees. You can create, work in, merge, and exit worktrees entirely from the chat UI.

NOTE The git worktrees integration is an experimental feature.

The worktrees integrations offers 3 slash commands:

Command	Description
/worktree <name>
Create a git worktree on branch <name> and move into it (skips creating it if it already exists)
/wt-merge [branch]
Merge the worktree branch into [branch] (default: main /master ), push, clean up, and return to the main repo
/wt-exit
Return to the main repo without merging

Create—/worktree feature-x

creates a new branch and worktree directory and moves you there.Work— Use zerostack normally; changes stay on the feature branch.** Merge**—/wt-merge

tells the agent to merge the branch, push, clean up, and return to the main repo.Exit—/wt-exit

immediately returns to the main repo without merging.

When you quit zerostack while in a worktree, the --wt-auto-merge

flag (or --parallel

, which implies it) causes zerostack to attempt merging the worktree branch before exiting.

Clean merge: completes silently (merge, push, remove worktree, delete branch).** Merge conflicts**: zerostack lists conflicting files and prompts:

[a]bort  [l]eave for manual resolution  [h]elp (agent resolves)

a

– abort the merge, restore clean state, do not delete the worktree.l

– leave the conflict state in the main repo for manualgit mergetool

.h

– abort the merge, then spawn the agent to redo the merge with interactive conflict-resolution guidance (same as/wt-merge

).

Flag	Description
--worktree <name>
Create a worktree on branch <name> and cd into it.
--wt-auto-merge
Auto-merge worktree branch on exit.
--parallel
Create a timestamped worktree with auto-merge on exit.
--wt-force
Force worktree remove and branch delete (-D ) even if dirty.
--wt-base-dir <dir>
Base directory for worktrees (default: parent of repo).

ACP is a JSON-RPC based protocol that standardizes communication between code editors (IDEs, text-editors, etc.) and coding agents. With the acp

feature enabled, zerostack acts as an ACP Agent server, allowing editors like Zed to connect to it as a coding agent backend.

NOTE: ACP support is gated behind the acp

feature and is not included in the default build.

zerostack --acp

zerostack --acp --acp-host 0.0.0.0 --acp-port 7243

In ~/.local/share/zerostack/config.yaml

:

acp_servers:
  my-editor:
    host: 127.0.0.1
    port: 7243

ACP mode requires setting up an LLM provider (the standard --provider

, --model

, and API key env vars apply). Without it, zerostack cannot process prompts.

• OpenRouter (default)
• OpenAI-compatible (vLLM, LiteLLM, etc.)
• Anthropic
• Gemini
• Ollama

Custom providers can be configured with any base URL and API key environment variable in ~/.local/share/zerostack/config.yaml

.

GPL-3.0-only

source & further reading

github.com — original article