Myco works alongside the coding agents, subagents, and agent teams you already use — giving them richer senses, durable project knowledge, semantic codebase awareness, and workflows that evolve with your code.
One install, every project. Myco connects to the coding agents on your machine, starts a local service, and registers git projects as you work.
22+).
Quickstart
Myco augments the agents you already use with richer senses, shared project knowledge, semantic codebase awareness, and learned workflows that stay current as the project changes.
Prompts, tool use, summaries, plans, screenshots, decisions, and pitfalls land in a local Grove vault. Myco keeps what still matters, supersedes stale observations, and turns repeated lessons into wisdom your agents can reuse.
The agent harness surveys your vault for procedural patterns with cross-session evidence, drafts a SKILL.md, and runs quality checks. You approve and publish — publishing is what writes it to .agents/skills/ for your agents. As the code changes, those skills can evolve too.
Join a teammate's Myco as your Team Host and every connected project's knowledge reaches team storage directly, machine to machine. A secure external endpoint exposes that knowledge to managed agents beyond your laptop.
The dashboard is how you work with Myco day to day. Configure providers and per-task models, manage projects and Groves, approve skill candidates, browse sessions and the project map, and watch the intelligence pipeline run — all on localhost, all yours, no account required.
Live view of concurrent agent work on your machine — prompts, tools, summaries, and extracted project knowledge. Click any session to walk its activity trail.
Approve or dismiss auto-generated skill candidates. Every change is recorded as lineage — what was in generation N, and why it's different in N+1.
Myco slows down when the project is idle and wakes back up when work resumes, so intelligence keeps moving without demanding attention.
Every session can produce durable observations. Myco turns repeated, evidenced patterns into skills your connected agents can follow on the next run.
The agent harness scans spores, sessions, and plans for procedural patterns with cross-session evidence. One-off observations don't qualify.
myco:survey≥ 0.7 qualityCandidates land in the Skills dashboard with rationale and confidence scores. You stay in the loop on what becomes canon for your project.
skills / candidateshuman-in-loopA reusable skill is drafted from the source material and validated; publishing it writes .agents/skills/ for your connected agents.
publishgen 1As your code evolves, so do your skills. Drift is detected, stale content rewritten, oversized skills split, outdated skills retired — with full lineage.
lineage recordedgen N+1Approved skills reach each connected agent in its native format, so every teammate can follow the workflows your best sessions established.
Your project's standard of excellence gets enforced, not just documented. The happy path becomes the default path.
Skills refresh as the code does. Each generation keeps enough lineage to understand why the workflow changed.
Cortex is Myco's synthesis layer — where accumulated project knowledge becomes context your agent can use. At session start it hands the agent a short project briefing; on demand it serves a deeper project digest; and Canopy keeps a living map of your codebase the agent can navigate before opening a single file.
One teammate's Myco becomes the team's shared home. Everyone else joins it at its own public address and connects the projects they want the team to share.
No wrappers, no forked CLIs, no ecosystem lock-in. Myco runs once on your machine and works alongside whatever agent you already use — adding shared project knowledge, codebase awareness, and learned workflows. Open any project with a supported agent and Myco connects to it automatically.
Guides cover everything from your first install to running a Team Host for your whole team.
Install Myco, open the dashboard, and configure providers. The supported path is macOS; Linux and Windows are in beta (Windows x64 only).
Read guide Getting startedKeep Myco current — how updates roll out, what changes between releases, and how to upgrade safely.
Read guide Getting startedMoving from OAK to Myco — what maps across, what's different, and how to bring your knowledge with you.
Read guide ServiceWhat gets captured, how the local service behaves, and what to do when something needs attention.
Read guide ProjectsOrganize projects into Groves, move or archive projects, and choose which intelligence capabilities each project runs.
Read guide SkillsThe four stages — survey, approve, generate, evolve — and the SKILL.md format Myco uses.
Read guide CortexCode intelligence for your agents — file anatomy before a read, a navigable project map, and search by what code does. How Myco cuts speculative reads.
Read guide TeamHost a team from your dashboard or join a teammate's, connect projects, and expose team knowledge to agents that aren't Myco members.
Read guide AdvancedThe background intelligence pipeline — phases, providers, task scheduling, and per-phase model overrides.
Read guide ReferenceFull reference for the MCP tools Myco exposes — search, recall, remember, plans, skills.
Read guide AdvancedHow Myco connects to each coding agent while preserving its native workflow.
Read guide AdvancedCoordinate multiple agents on one project — how teams share context and divide work across a Grove.
Read guideMyco is independent open source from Goondocks Consulting. Sponsorship helps fund maintenance, real workflow testing, documentation, release work, and the intelligence pipeline that keeps project knowledge evolving.
Myco runs entirely on your machine by default — no account, no vendor lock-in. Join a Team Host when you're ready to share knowledge across the team.