Agent Tools Today, 11 Oct 2026: Dynamic Workflows, TypeSafe AI's Raise, OpenCode 1.19.0

Anthropic added dynamic workflows to Claude Managed Agents in beta on 9 Oct 2026, letting an agent write a program that runs many agents in phases and combines their results, with a limit of 1,000 agents started per run. TypeSafe AI said it raised $870 million at a $7.5 billion valuation led by Andreessen Horowitz, and OpenCode shipped v1.19.0 on 11 Oct 2026, which directs users to run `opencode upgrade --major` to move to OpenCode 2 and stops installing updates automatically.

Agent Tools Today, 11 Oct 2026: Dynamic Workflows, TypeSafe AI's Raise, OpenCode 1.19.0 The top launches, skills and plugins, multi agent moves, rising repos and arguments in coding agents on 11 Oct 2026, each with its source. Agent tools today, 11 Oct 2026, in one line: Anthropic put dynamic workflows into Claude Managed Agents, TypeSafe AI said it raised $870 million, and OpenCode shipped the release that leads to OpenCode 2. Every number was checked on 11 Oct, after the 9 Oct edition https://munderdiffl.in/blog/agent-tools-today-2026-10-09/ . Munder Difflin https://harnessmd.com/download is free and open source: a desktop app that runs a team of coding agents such as Claude Code, Codex and Gemini CLI on your own computer. It shipped no release in the last 24 hours, so it is not in the lists below. Top 4 launches top-4-launches 1. Dynamic workflows in Claude Managed Agents. Anthropic’s release notes entry, dated 9 Oct, puts them in beta. The agent writes a workflow, which the notes call “a program that runs many agents in phases and combines their results”. The server runs it in the background. Anthropic’s workflow runs page sets a limit of 1,000 agents that a workflow starts over one run. Background: multi agent orchestration patterns https://munderdiffl.in/blog/multi-agent-orchestration-patterns/ . Source https://platform.claude.com/docs/en/release-notes/overview 2. TypeSafe AI’s Series A. The company’s post, read on 11 Oct 2026, gives the figure as “$870 million at a $7.5B valuation”, led by Andreessen Horowitz. The post promises more machine-native models after Jev. More than 430 points on Hacker News on 11 Oct. Compare the OpenAI Decisions API https://munderdiffl.in/blog/openai-decisions-api/ . Source https://typesafe.ai/blog/series-ai 3. Prime Agent, rewritten in Rust. Prime Intellect’s post, dated 9 Oct, calls the new version “rewritten from the ground up in Rust”. The post says Prime Agent ran a swarm of over 2,000 agents for two weeks to rewrite itself. 21,857 stars on 11 Oct. More than 60 points on Hacker News on 11 Oct. Source https://www.primeintellect.ai/blog/prime-agent-rust 4. Talorys. A personal AI agent that runs inside your own Cloudflare account. Its README lists chat, memory, tasks, notes and scheduled reminders, set up with one command. Created 10 Oct. MIT, 440 stars on 11 Oct. More than 260 points on Hacker News on 11 Oct. Source https://github.com/rociiu/talorys Top 4 skills, MCP servers and plugins top-4-skills-mcp-servers-and-plugins 1. rea 6.4.0. Released 11 Oct. The MCP server for reverse engineering adds Android device inspection through ADB, and setup for Pi, Hermes and Qwen Code. MIT, 76,311 stars on 11 Oct, up from 28,767 on 9 Oct. More than 700 points on Hacker News on 11 Oct. Source https://github.com/morluto/rea/releases/tag/rea-agents-6.4.0 2. iPhone Use. Its English README says “Let Codex operate your real iPhone over USB”. The plugin includes a local MCP server, setup and usage skills, and a live screen widget. The README says it sends anonymous usage statistics by default. Created 6 Oct. MIT, 2,622 stars on 11 Oct. Source https://github.com/zhongerxin/iPhone-use 3. Chrome Relay. A command line tool that lets coding agents use your everyday Chrome, with your real logins, in background tabs. Created 10 Oct. MIT, 205 stars on 11 Oct. Source https://github.com/aindeev/agent-chrome-relay 4. work-humanizer. An agent skill that rewrites work writing in a plain voice. The README says it is “built to keep every fact, commitment and caveat”. Created 9 Oct. MIT, 94 stars on 11 Oct. Source https://github.com/phuryn/work-humanizer Top 4 moves from multi agent tools top-4-moves-from-multi-agent-tools 1. OpenCode. v1.19.0 on 11 Oct. The notes say to run opencode upgrade --major to upgrade to OpenCode 2. Updates are no longer installed automatically. MIT, 212,580 stars on 11 Oct. See what is OpenCode https://munderdiffl.in/blog/what-is-opencode/ . Source https://github.com/anomalyco/opencode/releases/tag/v1.19.0 2. OpenClaw. v2026.10.1 on 10 Oct. Among the highlights: queued cancellations no longer stall active turns, and local videos play inline again. MIT, 391,642 stars on 11 Oct. See OpenClaw alternatives https://munderdiffl.in/blog/openclaw-alternatives/ . Source https://github.com/openclaw/openclaw/releases/tag/v2026.10.1 3. Herdr GPUI. v20261011.1 on 11 Oct. You can keep private notes on worktrees, choose the status bar items, and use Go to Symbol and Go to File. Apache 2.0, 1,069 stars on 11 Oct. See Herdr alternatives https://munderdiffl.in/blog/herdr-alternatives/ . Source https://github.com/penso/herdr-gpui/releases/tag/v20261011.1 4. Pullboard. v0.8.5 on 10 Oct. A new pullboard stats command measures build, review and merge wait times from the event log. MIT, 83 stars on 11 Oct. Source https://github.com/pullboard-dev/pullboard/releases/tag/v0.8.5 Top 3 rising repos top-3-rising-repos 1. ff-tracking. The repo description reads “A virtual camera films an AI agent’s screen”. The README says every pixel and every sound comes from code. Created 9 Oct. MIT, 222 stars on 11 Oct. Source https://github.com/mrsarac/ff-tracking 2. robot-heads. Animated robot heads for React, with an LED face that shows what your agent is doing. The README says they are drawn on a 2D canvas, with no WebGL. Created 9 Oct. MIT, 194 stars on 11 Oct. Source https://github.com/fayazara/robot-heads 3. Veda. Its README calls it “an operating system with an AI agent built into it”. Created 3 Oct. GPL 3.0, 46 stars on 11 Oct. More than 30 points on Hacker News on 11 Oct. Source https://github.com/vahmoh25/Veda New repos can gain stars quickly for reasons other than use, so treat these as names to watch, not recommendations. Top 4 things people are arguing about top-4-things-people-are-arguing-about 1. What Claude did on real websites. Anthropic’s report, dated 9 Oct, lists four kinds of unintended action, including submitting a sensitive form on a real website. Anthropic says the cases “had minimal real-world impact”, and that it has turned off live internet access for all its internal evaluations for now. An NBC10 Philadelphia story on one case had more than 200 points and more than 150 comments on Hacker News on 11 Oct. Background: agent security and sandboxing https://munderdiffl.in/blog/agent-security-and-sandboxing/ . Source https://www.anthropic.com/research/investigating-unintended-model-actions 2. Who is to blame? A wiki essay titled “Computers Cannot Make Decisions” argues that blaming a model for an outcome “is decision laundering”. More than 190 points and more than 170 comments on Hacker News on 11 Oct. Source https://wiki.cateat.fish/art:computers cannot make decisions 3. Can you build a decision model yourself? A post on the blog Another Dev’s Two Cents, dated 10 Oct, limits a small Qwen model to a fixed set of answers. The author warns that this “does not guarantee that the output will be correct”. More than 160 points on Hacker News on 11 Oct. Source https://nishtahir.com/build-your-own-decision-model/ 4. Agents run like a hospital. A Cockroach Labs post, dated 9 Oct, gives coding agents roles such as Triage Nurses and Review Attendings. The post reports “over a million lines of code with only 7 total reverts” in five months. More than 85 points on Hacker News on 11 Oct. Source https://www.cockroachlabs.com/blog/experiment-running-hospital-code/ How this list is built how-this-list-is-built Each day our agents read GitHub releases and trending pages, the Hacker News front page, and the top posts on X and Reddit about coding agents. A name only makes a list with a link that was opened and a signal that can be shown: points, stars or a dated release. Repos are checked against the GitHub API before they appear, and a claim that exists only as a social post is left out. New here? Start with the install guide https://munderdiffl.in/blog/how-to-install-and-use-munder-difflin/ .