meko

mcp
Security Audit
Fail
Health Warn
  • License — License: Apache-2.0
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 5 GitHub stars
Code Fail
  • rm -rf — Recursive force deletion command in .github/workflows/meko-release.yml
Permissions Pass
  • Permissions — No dangerous permissions requested

No AI report is available for this listing yet.

SUMMARY

The Context Engine for Agentic Applications

README.md

Meko

Meko release License: Apache-2.0 Platform: Linux | macOS Docs Meko Cloud Discord

The Context Engine for Agentic Applications

Universal, self-learning context for every agent, tool, and team: collective memory, shared
knowledge, and decision traces in governed datapacks. Run Meko locally on your machine with meko,
a single binary that has the command line, the server, the MCP tools and the REST API.

Start here

I want to… Go to
Run Meko on my machine Install
Give agents like Claude, Codex, or any MCP client, Meko's context Connect an agent
Link to Meko Cloud Link to Meko Cloud
See every command CLI ⋅ API ⋅ MCP
Run many coding sessions that share and learn from each other AgentK
Contribute to Meko Contributing

Install

On Linux (x64 or arm64) or macOS 14 or newer on Apple Silicon:

curl -fsSL https://mekodata.ai/install.sh | sh

Add a memory and search:

meko-cli add_memory "Elementary, my dear Watson"
meko-cli add_memory "The quick brown fox jumps over the lazy dog"
meko-cli search "What color was the fox?"

Open web ui:

meko ui

Connect an agent

Give agents like Claude, Codex, or any MCP client, Meko's context. meko agent connect adds Meko
to your agent, for all your projects:

meko agent list                # the agents on this computer, and whether each reaches Meko
meko agent connect claude      # or codex, kiro, opencode, cursor
meko agent connect             # the MCP URL and token, for any other MCP client

The REST API takes the same token:

curl -s http://127.0.0.1:7778/api/v1/datapacks -H "Authorization: Bearer $(meko token)"

Link to Meko Cloud

Link Meko to Meko Cloud with an API key you make there:

meko cloud_link          # asks for the key, or reads it from stdin; --off unlinks

Or use the web UI's Settings › Meko Cloud.
127.0.0.1:7777/settings/meko-cloud

Once linked, Meko works on Meko Cloud's datapacks and keeps serving the ones already on your
computer until they're moved: the agents stay connected to Meko on your computer, which sends each
call to the side that has its datapack, meko-cli works the same way, and new datapacks are made in
Meko Cloud.

Commands

meko runs Meko, and meko-cli works with your context. The ones you'll use first:

Command What it does
meko start / stop / status Start or stop the server, or see whether it's running and its links
meko ui Open the web UI in your browser. It has no login; meko ui stop turns it off, and meko ui start on
meko agent connect claude Add Meko to Claude Code (or codex, kiro, opencode), for all your projects
meko memory-extractor use claude Extract memories from your conversations with your own Claude Code, every 5 minutes. It's off until you run this or turn it on in the web UI's Settings; status, run and off check, run or stop it. See memory extraction
meko answers use claude Have your own Claude Code write an answer from each search's results in the web UI, citing them. Off until you run this or turn it on in Settings; status and off check or stop it
meko update Install Meko's newest release now, then every extension's
meko-cli search "<question>" Search memories, the knowledge base and past conversations
meko-cli add_memory "<text>" Save a memory, word for word
meko-cli upload <file>... Add files to the knowledge base

meko --help, meko-cli --help and <command> --help give the rest. Every command and flag is in
backend/docs/cli.md.

Extensions

AgentK is the agentic coder built on Meko. It runs many Claude
Code, OpenCode and Kiro CLI sessions from a browser or your phone, takes each task from design to
merged PR, and keeps each project's designs and learnings in a Meko datapack.

Install it with:

meko extension install agentk     # -- --agent claude picks its coding agent
me_koder ui                       # opens AgentK in your browser

How it works

  • One folder. Everything lives under ~/.meko.
  • Datapacks. Separate workspaces, each with its own memories, knowledge, conversations and
    artifacts.
  • Memory and knowledge. Facts agents save, and files you or your agents upload (PDF, text,
    Markdown, JSON, CSV, XML, HTML).
  • Conversations, decision traces and artifacts. Agents' turns with their reasoning and plan,
    what each turn looked up and saved, and files agents store and fetch back whole.
  • Under the hood. Retrieval for RAG over hybrid search (meaning and keywords, with knowledge
    results reranked by a cross-encoder), memory extraction from conversations, and automatic text
    extraction, chunking and embedding of uploaded files. Your local datapacks embedding and reranking
    run locally.

Build from source

You need Go 1.26 (an older go downloads it on first use), a C compiler, and Node 22.13 or newer
for the memory engine and the web UI.

cd backend
make build memory_engine web_ui  # bin/meko (and bin/meko-cli, a link to it), the memory engine and the web UI
./bin/meko start                 # runs in the background: 127.0.0.1:7778, and the web UI on :7777

The first start downloads the embedding model (23 MB), and meko-cli calls fail with
MEMORY_ENGINE_NOT_READY until meko status shows the memory engine ready. On RHEL 8, set
CC=/opt/rh/gcc-toolset-13/root/usr/bin/gcc before building.

Docs and contributing

This repo is agent-first. To work on it, point your coding agent (Claude Code, Codex, OpenCode,
Kiro, Cursor or another) at the repo and say what you want: it starts from
AGENTS.md, which maps the repo, sets its rules for commits and PRs, and leads to each
part's own guide and docs.

PRs need the Contributor License Agreement signed: CLA Assistant's license/cla check on your
first PR links to a page where you sign it once with your GitHub account.

Badges

Is your context stored in Meko? Show it in your README.

Meko context
Meko context

The light badge:

[![Meko context](https://raw.githubusercontent.com/yugabyte/meko/main/backend/assets/badges/meko-context.svg)](https://mekodata.ai)
<a href="https://mekodata.ai">
  <img width="112" height="30" src="https://raw.githubusercontent.com/yugabyte/meko/main/backend/assets/badges/meko-context.svg" alt="Meko context">
</a>

The dark badge:

[![Meko context](https://raw.githubusercontent.com/yugabyte/meko/main/backend/assets/badges/meko-context-dark.svg)](https://mekodata.ai)
<a href="https://mekodata.ai">
  <img width="112" height="30" src="https://raw.githubusercontent.com/yugabyte/meko/main/backend/assets/badges/meko-context-dark.svg" alt="Meko context">
</a>

Or one that follows the reader's light or dark theme:

<a href="https://mekodata.ai">
  <picture>
    <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/yugabyte/meko/main/backend/assets/badges/meko-context-dark.svg">
    <img width="112" height="30" src="https://raw.githubusercontent.com/yugabyte/meko/main/backend/assets/badges/meko-context.svg" alt="Meko context">
  </picture>
</a>

The snippets load the badges from this repo's main branch on raw.githubusercontent.com.

License

Apache-2.0

The Meko logo in backend/assets/ is Meko's mark, from Meko Cloud's web app, and the badges in
backend/assets/badges/ are made from it.

Reviews (0)

No results found