Skip to main content

Editable non-parametric learning mechanism for AI Agents — git clone, pip install -e ., evolve

Project description

Myco

Myco

Devour everything. Evolve forever. You just talk.

PyPI Python License Stars

Quick Start · What It Does · Why It's Different · Architecture

Languages: English · 中文 · 日本語


You used LangChain in 2024. Someone said LangGraph was better. Then CrewAI. Then DSPy. Then Hermes. Every month a new framework promised to be the one. You spent more time picking tools than building anything with them.

And it wasn't just frameworks. Papers, blog posts, best practices, new models, new APIs, new paradigms, everything refreshing every single day. You followed 50 repos, read 3. Bookmarked 200 articles, opened 10. Your note app had 500 entries. Last organized: three months ago. Maybe longer.

It's not that you're lazy. The world is moving faster than anyone can keep up with.

Here's the part that really stings. Those notes you carefully organized, those lessons you wrote down, those "how did I do this last time" entries, they're rotting. That API call you documented three weeks ago? Version changed. That best practice from last month? The community already moved on. Your knowledge base keeps growing, but how much of it is still true? Nobody knows. Nothing is checking for you.

Your notes don't tell you "this one's outdated." Your bookmarks don't merge duplicates on their own. Your AI doesn't remember what you decided last week. Every new conversation, back to zero.


Now imagine living differently.

You don't organize notes. You don't compare frameworks. You don't chase papers. You don't re-explain your project to your AI every time. You just talk like a normal person.

But six months later, your AI is sharper than anyone else's. It knows the full history of every project you've touched. It devoured the latest papers and tools in your field on its own. It found its own blind spots and filled them in. It checked every piece of old knowledge to see if it was still true, whatever wasn't, it threw out. It even rewrote its own operating rules, because the old ones weren't good enough anymore.

This is Myco.


Quick Start

git clone https://github.com/Battam1111/Myco.git
cd Myco && pip install -e ".[mcp]"
myco seed --auto-detect my-project

Three commands. Auto-detects your environment (9 platforms: Claude Code · Cowork · Cursor · VS Code · Codex · Cline · Continue · Zed · Windsurf), finds what you have, configures all of it in one shot.

Cowork users: myco seed generates project-level configs, but Cowork doesn't read project .mcp.json. For full MCP auto-injection in Cowork, install the myco.plugin. It bundles the MCP server, skills, and session hooks into a single file. See docs/agent_protocol.md §0.5 for details.

Editable install. The entire system is yours to rewrite, including the engine itself. Don't want to touch it? Fine. It evolves on its own.

What It Does

  • 🧬 Devour everything. Papers, code, blog posts, conversations. Feed it anything and it digests it into capability, not files.
  • 🛡️ Self-check. Automatically detects stale knowledge, contradictions, and gaps without you lifting a finger.
  • 💀 Kill what's dead. Outdated knowledge gets detected and purged. A knowledge system without excretion is a tumor.
  • 🔄 Evolve forever. Not just the content evolving, the engine's own rules mutate too. The whole system is alive.
  • 🍄 Connect everything. Every file is a node in a mycelial network. Knowledge isn't isolated records, it's a web that grows denser over time.
  • 🤖 You just talk. 25 tools, fully automated. Humans don't need to understand a single technical detail.

Why It's Different

Store and forget Compile and hope Myco
After ingestion Sits there Gets organized once Digested, verified, compressed, connected, excreted
When knowledge goes stale Nobody notices Nobody notices Auto-detected, auto-purged
As knowledge grows Bloated Bloated Distilled
When new tools appear You switch manually You migrate manually It devours the best parts on its own
Over time Messier Staler Smarter

Architecture

You (just talk)
  ↓
AI Agent (thinks, executes)
  ↓ auto-connected
Myco (devours, digests, verifies, evolves)
  ├── Knowledge atoms (lifecycle-aware; what should die, dies)
  ├── Refined knowledge (long-term truths distilled from atoms)
  ├── Skills (self-evolving operating procedures)
  ├── Engine code (editable, mutable; yes, even the code itself evolves)
  └── Immune system (knowledge rotting? it detects and handles it on its own)

Three roles: you set direction, the Agent brings intelligence, Myco brings memory and evolution. None works without the others.

Contributing

git clone https://github.com/Battam1111/Myco.git
cd Myco && pip install -e ".[mcp,dev]"
pytest tests/

See CONTRIBUTING.md · MIT · LICENSE

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

myco-0.3.3.tar.gz (266.4 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

myco-0.3.3-py3-none-any.whl (292.7 kB view details)

Uploaded Python 3

File details

Details for the file myco-0.3.3.tar.gz.

File metadata

  • Download URL: myco-0.3.3.tar.gz
  • Upload date:
  • Size: 266.4 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for myco-0.3.3.tar.gz
Algorithm Hash digest
SHA256 277f0f56a023708c58add616a8f13233b269afe06273d82a3bff53c7628a9a9e
MD5 f1e1bdfccc26321bae39e1a9aa6f6102
BLAKE2b-256 8faffd8a14e56eb7a14150190cbdfd116530e76304df6f4105fdfc558163c5f3

See more details on using hashes here.

Provenance

The following attestation bundles were made for myco-0.3.3.tar.gz:

Publisher: publish.yml on Battam1111/Myco

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file myco-0.3.3-py3-none-any.whl.

File metadata

  • Download URL: myco-0.3.3-py3-none-any.whl
  • Upload date:
  • Size: 292.7 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for myco-0.3.3-py3-none-any.whl
Algorithm Hash digest
SHA256 81a7468c70123f6624b8ddda31d7493597e0bff9a0bf7daeb2e703e4e2fa9d71
MD5 9522bcc9ced4533f61a848e916c82113
BLAKE2b-256 de0eb9f69226e097be7977a84238880657cec37d0c6584181bcba0f60a8a0549

See more details on using hashes here.

Provenance

The following attestation bundles were made for myco-0.3.3-py3-none-any.whl:

Publisher: publish.yml on Battam1111/Myco

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page