Myco Quick Start
Myco is the nervous system for AI-assisted software teams. It captures session knowledge — events, observations, decisions, trade-offs, codebase context, and learned workflows — into a local Grove vault and routes the right context back to the coding agents, subagents, and agent teams you already use.
Myco does not replace an agent’s reasoning, native memory, tools, or workflow. It augments those systems with shared project knowledge, semantic codebase awareness, and workflows that compound across sessions and teammates.
Requirements
- Node.js 22+
- At least one supported coding agent — Claude Code, Cursor, Codex, Copilot, Google Antigravity, Windsurf, OpenCode, or Pi
- macOS for the supported path in the current release. Linux and Windows packages are available for early testing, but they are experimental.
Provider configuration is optional at install time. Myco captures sessions and provides full-text search immediately. To enable spores, digests, semantic search, Canopy summaries, and skill lifecycle features, configure intelligence and embedding providers in the dashboard after install.
When you’re ready to enable intelligence features, you’ll need:
- Myco Agent provider (one of):
- Embedding provider (one of):
- Ollama with
bge-m3model (local, free, recommended) - OpenAI-compatible endpoint
- Ollama with
Install
curl -fsSL https://myco.sh/install.sh | sh
myco open
On Windows (PowerShell, experimental):
irm https://myco.sh/install.ps1 | iex
Or install manually with npm:
npm install -g @goondocks/myco
Upgrade Existing Installs
Existing users upgrade the main product the same way:
npm update -g @goondocks/myco
That updates the local CLI, service, dashboard, agent connections, and built-in intelligence pipeline. If you later install one of the standalone operator CLIs, the Operations page will detect and apply updates for those installed Myco packages too. You do not need to install extra packages unless you want one of the standalone operator CLIs:
@goondocks/myco-teamfor direct team-worker administration commands@goondocks/myco-collectivefor cross-project Collective administration
That’s it — ready by default
Once Myco is installed, the local service starts automatically. On first boot it creates a default Grove to hold your captured sessions, then connects detected coding agents to Myco. The first time you use a supported agent from a real git repo, Myco auto-registers that project into your default Grove. The dashboard is available immediately — you can configure intelligence providers, create additional Groves, or move projects between Groves before your first captured session.
New agent installed later? Myco detects it automatically, or you can re-detect from the dashboard’s Symbionts page.
To share a project with teammates, commit the .myco/project.toml file — Myco creates it automatically and tracks it in git. It carries the project’s identity, so when a teammate clones the repo their sessions land in the same project. Everything else under .myco/ is per-machine and stays out of git.
Myco captures inside git repositories only — a project is a git repo. Work in a non-git folder isn’t recorded; open a git repo (or run git init) to start capturing.
Session capture starts immediately and you get full-text search out of the box. To enable the deeper intelligence features, configure an agent provider in the dashboard’s Myco Agent section.
Configure Providers in the Dashboard
Open the dashboard to the Settings page. Two cards are at the top:
- Myco Agent — pick Anthropic, Ollama, or LM Studio. The dropdown lists detected models. Click Save to enable the intelligence pipeline.
- Embedding — pick Ollama or an OpenAI-compatible endpoint. Embedding models are filtered automatically.
Each section saves independently. You can enable just one (e.g. embedding-only for semantic search) or both.
Pull Ollama Models (if using Ollama)
ollama pull bge-m3 # for embeddings
ollama pull granite4:small-h # for the agent (any LLM works)
Health Check
After setup, verify everything is connected:
myco doctor
Doctor warns rather than errors when provider config is absent, because capture and full-text search are useful before intelligence providers are configured.
What Happens Next
Once installed, Myco works automatically:
- Session start: Myco adds a project briefing when intelligence providers are configured
- During the session: Activity (prompts, tool calls, responses) is captured in the vault
- After the session: Myco extracts spores, generates summaries, and maintains the knowledge graph
- Over time: Repeated lessons become wisdom, stale knowledge can be superseded, and approved skills evolve with the code
You don’t need to drive the loop manually. Myco captures knowledge in the background and augments your agents without taking over their native memory or workflows.
Dashboard
Myco includes a local web dashboard for configuration and operations management. After the service starts, open it with:
myco open
You can also open the dashboard directly at http://localhost:20915/. If your local install reports a different service URL, myco open will open the right one.
The dashboard lets you:
- Configure intelligence providers, per-task model assignments, and embedding settings
- Manage Groves, projects, archive/unarchive, project capability toggles, and destructive delete with confirmation
- Run operations like intelligence agent runs, index rebuilds, and manual digest cycles
- Monitor service health, background activity, and system stats
- View logs in real-time with level filtering
The Settings page shows where each setting applies. Changes take effect automatically, and the dashboard prompts for a restart only when one is required.
See Grove Management for the guide to Groves, project movement, capture-only projects, and capability toggles.
MCP tools
Myco exposes a set of MCP tools to your coding agent — search, recall, remember, browse sessions and plans, traverse the knowledge graph, and inspect skills. Agents discover them automatically through MCP.
See Agent Tools for the full reference.
Skills
Myco auto-generates project-specific skills from accumulated vault knowledge — repeatable workflows that teach every connected agent how to work in your codebase. As the intelligence pipeline processes sessions, it identifies procedural patterns with cross-session evidence and surfaces them as candidates. You approve what becomes canon in the Skills dashboard, and Myco writes validated SKILL.md files to .agents/skills/. Skills evolve automatically as your code does.
Myco also ships two slash-command skills out of the box:
| Command | What it does |
|---|---|
/myco |
Guidance on using Myco during ongoing work — design decisions, debugging, vault hygiene |
/myco-rules |
Keep AGENTS.md minimal, durable, and canonical across agents |
See the Skills docs for the full auto-curation lifecycle.
Troubleshooting
Something not working?
Run the health check:
myco doctor
To auto-repair fixable issues:
myco doctor --fix
Service not starting
Myco is installed as a per-user service and starts at login. Linux and Windows service support is experimental. If it fails:
myco restart # Manual restart
myco stats # Check status
myco logs # Tail the service log
If Myco is running but no projects are registered yet, start work with a supported agent from any git-tracked project. Myco auto-registers that project into your default Grove and capture starts immediately.
No observations being captured
Verify your intelligence provider is configured and reachable:
myco doctor
For local providers, ensure they’re running:
# For Ollama
curl http://localhost:11434/api/tags
# For LM Studio
curl http://localhost:1234/v1/models
Uninstall
myco remove # remove Myco's contributions from every agent's global config
myco remove --purge # additionally remove ~/.myco/ itself
myco remove preserves user-pre-existing keys in shared agent config files (for example, a [features].hooks entry you added to ~/.codex/config.toml yourself).
Optional Operator CLIs
Most users only need @goondocks/myco. Install the extra CLIs only when you want their dedicated operator surfaces.
Standalone team CLI
npm install -g @goondocks/myco-team
Operators only — teammates join a team from the dashboard’s Team page without this CLI. Use it for:
myco-team install --name "<team name>"— provision a team (deploys its Worker, prints a Worker URL + Team key)myco-team update --team-id <id>— redeploy the team’s Workermyco-team status --team-id <id>— show Worker info and credentialsmyco-team rotate-tokens --team-id <id>— rotate the Team key and/or MCP tokenmyco-team reindex-vectors --team-id <id>— rebuild the team’s search indexmyco-team destroy --team-id <id>— tear down the team’s Cloudflare resources
Collective CLI
npm install -g @goondocks/myco-collective
Use it to create and manage a Myco Collective.