Install agents, not frameworks.
Open-source runtime for specialist AI agents. One command. Any model. Any domain.

Documentation · Tap Registry · Website

• The Problem
• Three Pillars
• Pillar 1 — Zero Config, Full Flexibility
• Pillar 2 — Sessions That Stay Sharp at Hour 4
• Pillar 3 — Cost as a Control Plane
• Quick Start
• How It Works
• Power Users — Roles, Workflows, Layers
• Embedders — ACP, WebSocket, Daemon
• Installation
• Configuration
• Architecture
• Contributing
• License

Building a specialist AI agent in 2026 means stitching together three different tools, writing glue code nobody wants to own, and praying it holds. You spend 45 minutes wiring MCP servers, system prompts, tool configs, and credentials — every domain, every machine, every time.

• Config Wars. No central registry. Skills here, MCP servers there, agent configs nowhere. The community calls it "solidarity in frustration."
• Generic AI hallucinates in expert domains. ChatGPT writes wrong drug dosages. Lawyers cite cases that don't exist. Multi-agent specialization is now the default architecture for serious work.
• One generic assistant for every task. Rust debugging, blood-test interpretation, contract review — same prompt, same tools. Drift compounds.
• Sessions break at hour 4. Naive truncation drops the decisions you need. Quality collapses. You restart.
• Bills surprise you. Cursor users posting $7K daily overages. No per-task budget, no kill switch.

Octomind ships specialist agents ready to run — and a runtime that grows with you.

Most agent tools force a tradeoff: zero-config (Lindy, no-code) or fully customizable (Mastra, LangGraph). Octomind gives you both. octomind run lawyer:sg works out of the box. Need a different model, a custom MCP server, a pipeline of agents — all live in TOML, no framework code.

octomind run developer            # general dev, language skills auto-activate
octomind run doctor:blood         # blood-test interpretation specialist
octomind run doctor:nutrition     # nutrition specialist

→ Fetches the agent manifest from the tap registry
→ Installs required binaries automatically (skips if already present)
→ Prompts once for any credentials, saves permanently
→ Spins up the right MCP servers for this domain
→ Loads specialist model config, system prompt, tool permissions
→ Ready in ~5 seconds, not 45 minutes

This is packaged expertise — not a prompt file, not a skill injection. The full stack, configured by the community, ready to run.

Every agent has built-in power tools that let it acquire new capabilities and spawn sub-agents mid-session, without restart:

User: "Cross-reference our Postgres metrics with the deployment log"

Agent:
  → mcp.enable(postgres-mcp)        # auto-detected need, no user prompt
  → agent.spawn(log_reader)         # delegates log parsing
  → results merge mid-session
  → mcp.disable(postgres-mcp)       # cleans up
  → presents the analysis

Most agent harnesses pre-load every available tool into context. Octomind starts focused for the domain and grows only when needed. Smaller context, lower cost, faster responses, no surprise tools.

octomind tap yourteam/tap                 # clones github.com/yourteam/octomind-tap
octomind tap yourteam/internal ~/path     # local tap for private agents

octomind run finance:analyst              # available immediately
octomind run security:owasp

Each tap is a Git repo. Each agent is one TOML file. Pull requests are contributions.

Want to publish your expertise? A doctor:medications, a lawyer:us, a devops:terraform. One file, and everyone with that problem gets a specialist instantly. How to write a tap agent →

Every coding agent degrades after a few hours. Context fills. Decisions get truncated. The agent forgets why it started.

Octomind's adaptive compaction engine runs automatically:

• Cache-aware — calculates if compaction is worth it before paying for it. Never breaks the prompt-cache hit by accident.
• Pressure-tiered — compacts more aggressively as context grows.
• Structurally preserving — keeps decisions, file references, errors, dependencies; drops noise.
• Plan-aware and free-form-aware — works whether you use the plan tool or have a free-form chat.
• Fully automatic — you never think about it.

The second-order benefit: smaller context means fewer tokens, faster responses, lower cost every turn after compaction fires. The three pillars compound.

Work on a hard problem for 4 hours. The agent still knows what it decided in hour one.

Pick the right model for each step. A cheap one for routine research, a frontier one for review — per-role, per-step, mid-session swap. Real-time cost tracking and hard spending caps come for free.

# Per-role model selection — pay Opus only where it's worth it
[[roles]]
name = "researcher"
model = "openrouter:google/gemini-2.5-flash"   # cheap broad context

[[roles]]
name = "reviewer"
model = "anthropic:claude-opus-4-7"            # precision where it counts

# Hard spending limits — enforced, not advisory
max_request_spending_threshold = 0.50    # USD per request
max_session_spending_threshold = 5.00    # USD per session

• Per-role and per-layer model selection across many providers via octolib — different roles can run on different vendors.
• Mid-session model swap with /model anthropic:claude-haiku-4-5.
• Real-time cost tracking per request and per session.
• Cache-aware token accounting (cache_read_tokens, cache_write_tokens separated from input/output).
• Hard spending thresholds with enforcement — agent stops, falls back, or warns before the bill.

Cursor users get $7,000 surprise bills. Octomind agents trip a budget and stop, fall back, or warn — before the bill, not after.

# Install (macOS & Linux)
curl -fsSL https://raw.githubusercontent.com/muvon/octomind/master/install.sh | bash

# One API key gets you many providers
export OPENROUTER_API_KEY="your_key"

# Start with a specialist — no setup required
octomind run developer

That's it. You're in an interactive session with a specialist that can read your code, run commands, edit files, and grow capabilities as needed.

view, text_editor, batch_edit, shell, semantic_search, structural_search, workdir — file operations come from the companion octofs MCP server, included by default in tap formulas that need them.

memorize, remember, forget, knowledge, relate, memory_graph — long-term memory, knowledge indexing, and relationship graphs. The companion octobrain MCP server is included by default in taps that need persistent context across sessions.

Octomind supports many providers — OpenRouter, OpenAI, Anthropic, Google, DeepSeek, Amazon Bedrock, Cloudflare, and more — via octolib. New providers added there become available in Octomind automatically. See Providers & Models for the current list and supported models.

Switch providers mid-session with /model anthropic:claude-sonnet-4-6. Mix providers across roles — cheap model for research, best model for execution. Cost tracked separately per provider.

For most users, taps are enough. For teams and power users, the configuration system is deep — all TOML, no code.

# Per-role: independent model, temperature, MCP servers, tools, system prompt
[[roles]]
name = "senior-reviewer"
model = "anthropic:claude-opus-4-7"
temperature = 0.2
[roles.mcp]
server_refs = ["filesystem", "github"]
allowed_tools = ["view", "ast_grep", "create_pr"]

# Workflows — multi-step, each step its own model and toolset
[[workflows]]
name = "deep_review"
[[workflows.steps]]
name = "analyze"
layer = "context_researcher"     # gemini-flash, broad context
[[workflows.steps]]
name = "critique"
layer = "senior_reviewer"        # claude-opus, precision

# Sandbox — lock all writes to current directory
sandbox = true

• Roles — model, temperature, system prompt, MCP servers, tool permissions per role.
• Layers — chained AI sub-agents that run after each response.
• Pipelines — deterministic script-driven pre-processing.
• Workflows — multi-step orchestrated task runners with validation loops.

See Configuration Reference for everything.

Octomind isn't just a CLI. It runs in every context an agent needs to live in:

# ACP — drop into any multi-agent system as a sub-agent
octomind acp developer:general

# Non-interactive — single message, plain output
octomind run developer "Explain the auth module" --format plain

# Structured JSON output for pipelines
octomind run developer "List TODO items" --schema todos.json --format jsonl

One binary. Every workflow.

curl -fsSL https://raw.githubusercontent.com/muvon/octomind/master/install.sh | bash

Detects OS and architecture, installs to ~/.local/bin/. macOS and Linux supported. Single Rust binary, no runtime dependencies.

cargo install octomind

Requires Rust 1.82+. See Building from Source.

git clone https://github.com/muvon/octomind.git
cd octomind
cargo build --release

# OpenRouter — recommended, access to many providers with one key
export OPENROUTER_API_KEY="your_key"

# Or any specific provider
export OPENAI_API_KEY="your_key"
export ANTHROPIC_API_KEY="your_key"
export DEEPSEEK_API_KEY="your_key"

Add to ~/.bashrc or ~/.zshrc for persistence.

octomind --version
octomind config       # generate default config
octomind run          # start your first session

Config lives at ~/.local/share/octomind/config/config.toml.

octomind config --show          # view current config
octomind config --validate      # validate config

Key areas:

• Roles — model, temperature, system prompt, MCP servers, tool permissions
• Workflows — multi-step AI processing with validation loops
• Pipelines — deterministic script-driven pre-processing
• MCP Servers — external tools and capabilities
• Spending Limits — per-request and per-session thresholds

Full reference: Configuration Reference.

Full list: Session Commands.

CLI / WebSocket / ACP / Daemon
            |
            v
       ChatSession                  <- src/session/chat/session/
            |
            +-- Roles               <- src/config/roles.rs
            +-- Layers              <- src/session/layers/
            +-- Pipelines           <- src/session/pipelines/
            +-- Workflows           <- src/session/workflows/
            +-- Learning            <- src/learning/
            +-- Adaptive compaction <- src/mcp/core/plan/compression.rs
            |
            +-- MCP servers         <- src/mcp/
                  +-- core/    plan, schedule, tap, capability
                  +-- runtime/ mcp, agent, skill
                  +-- (filesystem via external octofs)
                  +-- (brain via external octobrain)
                  +-- agent/   agent_* tools route tasks to layers

Config is the single source of truth. All defaults live in config-templates/default.toml. The resolved config drives everything: which model, which MCP servers, which layers, which role. No hardcoded values in code.

See Architecture for internals.

The most impactful contribution isn't code — it's specialist agents.

Every domain expert who publishes a specialist makes Octomind useful for an entirely new audience. A cardiologist publishing doctor:medications. A tax attorney publishing lawyer:us. A security researcher publishing security:owasp. One TOML file — and everyone with that problem gets a specialist-grade AI instantly.

• How to write a tap agent
• Open issues
• Building from source

• Installation & Setup
• Quickstart
• Configuration
• Providers & Models
• Sessions & Compression
• Roles
• MCP Tools
• Workflows
• Pipelines
• Learning
• WebSocket & ACP
• CLI Reference
• Config Reference

Full index: doc/README.md.

Apache License 2.0 — see LICENSE.

Octomind by Muvon | Website | Documentation