Skip to main content
AdOpenFree logoPromote your productReach more potential users and drive product growth and revenue.Advertise
OpenFree logo

Octomind

Free Listing

Octomind is an open-source AI coding agent and agent runtime: one binary, any model, MCP-native. It runs in the terminal, CI, or as a daemon, with portable agents defined in TOML and policy in guardrails.

Visit Muvon/octomind

Overview

Octomind is an open-source coding agent harness and agent runtime. It gives LLMs tools through MCP to read and edit files, run shells, search code, and delegate to sub-agents. The same session engine runs as an interactive CLI, a stdin pipe, a daemon, a WebSocket server, or an ACP agent. Policy is written as deterministic TOML guardrails, and a supervising model plans and checks the work so sessions can run unattended.

Key Features

  • Deterministic TOML guardrails provide pre-call guards, post-result hooks, and post-turn validators.
  • Any supported model can be used and swapped mid-session, including OpenRouter, Anthropic, OpenAI, Google, DeepSeek, Bedrock, and Ollama.
  • An agent is one TOML file containing models, tools, prompts, and guardrails that can be read, forked, and committed.
  • The runtime exposes the same harness through interactive CLI, non-interactive structured output, daemon plus send, WebSocket server, and ACP.
  • Adaptive compression preserves decisions, errors, and file references while dropping noise, supporting sessions that resume across days.
  • Tools remain dormant and activate by meaning, limiting context to the relevant tools rather than preloading many.

Use Cases

  • Running daily coding sessions in the terminal with packaged specialists such as developer:general.
  • Using non-interactive JSONL or plain output in CI, shell scripts, and scheduled jobs.
  • Keeping a background agent alive for days and injecting new work with the daemon and send commands.
  • Chaining specialists into multi-step workflows for research, writing, review, and polishing.

Getting Started

  • Install with the provided one-line script for macOS or Linux, the PowerShell command for Windows, Homebrew on macOS, Cargo, or a source build.
  • Run octomind once in a terminal to choose where models come from: Octomind hub, your own provider, or shipped defaults.
  • In a project, run octomind run developer:general to start a session with a packaged specialist.

Deployment & Requirements

  • The runtime is distributed as a single binary for macOS, Linux, and Windows.
  • The runtime works standalone with your own provider keys and does not require Octomind servers; signing in is a convenience.
  • Cloud execution is available as an optional Octomind Cloud offering.

Before You Adopt

  • License: Apache-2.0. Review its terms before using, modifying, or distributing the project.
  • Benchmark numbers are published historical results for a specific date, not a measurement of the current checkout.
  • API keys are intentionally not accepted in the config file and must come from the environment or a .env file.
  • The first run of a specialist may install its tools and request credentials.

Comments

Sign in to leave a comment.

More like Octomind

Favicon of dsh-superagent

dsh-superagent

Free ListingStars: 2.6K

A proactive coding agent for one GitHub repository

Coding Agents

Proactive coding agent for one GitHub repository, packaged for DeepSeek Harness. It listens to GitHub and optionally Feishu, opens work items, works independently, replies in the original conversation, and shows human-needed work in a web

Favicon of Coucou

Coucou

Free ListingStars: 6.8K

A tiny friend that lives in your notch

Coding Agents

Coucou is an open-source desktop companion that watches AI coding agents, lets you approve requests from the notch or phone, and adds chat, dictation, call notes, and service integrations.

Favicon of Crit

Crit

Free ListingStars: 1.3K

Line-precise review for AI coding agents

Coding Agents

Crit is an open-source, local-first review surface for AI coding agents. It renders plans, diffs, running apps, and HTML artifacts in the browser so you can comment on exact lines and let the agent revise.