MemForge plugin
A memory plugin for Claude Code and Codex: team knowledge from wikis, tickets and chat with citations and stale-fact checks.
- 103 ★ on the repository
- Last commit
Extended access
Why this level
- The server reads documents from connected sources and sends their text to the chosen model provider.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add plugins/memforgeDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The server reads documents from connected sources and sends their text to the chosen model provider.
This is third-party code. Review the repository files before installing.
What it does
MemForge syncs team sources such as Confluence, Jira, Teams and GitHub ahead of time and builds one memory from them. A model turns documents into precise facts and decisions with their conditions and exceptions. Agents search them over MCP by keyword and, with an embedding model connected, by meaning. Results carry no document text: a full document enters the context only when the agent asks for it. Every fact cites the passage behind it and, where possible, links to the original page, ticket or file. On each sync, on demand or on a schedule, changed documents are rechecked: a fact is kept, replaced or retired, and when two documents disagree the agent gets both with a conflict note. The open-source edition runs on your machine through Docker Compose and the plugin installs into Claude Code and Codex.
Who it is for. Teams whose knowledge sits in wikis, tickets and chat and whose coding agent needs decisions and conventions before editing.
Good fit when
- A new hire asks the agent to fix a ticket while the wiki records decisions that are not in the code.
- Two documents disagree and the agent must show both with dates.
- Each fact the agent relies on has to be checked against its source.
- Team knowledge should refresh on a schedule rather than be copied by hand.
Not a fit when
- You need shared memory for a whole team: the open edition is for one person on one machine.
- The server cannot run without Docker Compose and a model key.
- You need a source outside the list: you would have to write the connector yourself.
Example request
Before editing, check MemForge for decisions and conventions about payments and show the source of each one you rely on.Limitations
The open edition is single-user, has no accounts or authentication and listens only on `127.0.0.1`, so it must not be exposed to a network. The Team Cloud Edition is still on a waitlist. Docker with Compose v2 and an OpenRouter, OpenAI or Anthropic key are required, and without a generation model a sync stores content but creates no memory. Source text goes to the model providers you choose. The project is in beta: APIs, storage formats and plugin packaging may change, and an upgrade with a migration saves a recovery point. Model extraction can miss useful knowledge, so important claims are checked against the citation.
How to disable. Disable the plugin in Claude Code through `/plugin` or the Codex command and stop the server with `docker compose down` in the repository folder. Data is stored locally and is removed with the volumes.
FAQ
How do I install the plugin in Claude Code?
In a session run `/plugin marketplace add dodo-labs-ai/mem-forge`, then `/plugin install memory@memforge`. For Codex: `codex plugin marketplace add dodo-labs-ai/mem-forge` and `codex plugin add memory@memforge`.
How do I start the server?
Clone the repository, copy `.env.example` to `.env` and run `docker compose up --build`. The interface opens at `http://localhost:5174`.
Which sources can I connect?
Confluence, Jira, Teams, GitHub and local folders. Obsidian is planned and missing sources can use the connector interface.
How does the agent trust the facts it finds?
Each fact shows a citation and a source link. If the source no longer supports a fact, the fact is retired.
Do keys pass through the agent?
No. The agent starts the server and the plugin and you enter the model key yourself in the admin panel.
Discussion
No questions or reviews yet.
Related
PluginsPersistent memory for coding agents: insights are stored as markdown cards and recalled at the start of the next session.
A long-term memory plugin that keeps decisions, preferences and task state across sessions and injects only what a query needs
Shared semantic memory for agents: conversations are saved as markdown and recalled by search across Claude Code, Codex and other agents
A local workbench for Claude Code, Codex, Pi and OpenCode: memory changes wait for your review, then an audit follows each turn
Coding memory across sessions, machines and teammates: decisions and dead ends flow into later prompts on their own.
Knowledge-graph memory for agents: Claude Code and Codex plugins, an MCP server and a Python library
README in short
The README covers why it exists, five setup steps, a flow from sources to the agent and the open edition's limits. Details sit in the quickstart, architecture and release docs.