<h1><img src="https://raw.githubusercontent.com/dodo-labs-ai/mem-forge/main/.github/assets/logo.svg" width="36" align="absmiddle" alt=""> MemForge</h1>

**Give every agent your team's brain on day one.**

The memory layer for AI agents. Everything your team knows. Cited, and always
up to date. Built from your docs, tickets and chat, plus what you ask your
agents to remember.

<p>
  <a href="https://github.com/dodo-labs-ai/mem-forge/actions/workflows/ci.yml"><img alt="CI" src="https://github.com/dodo-labs-ai/mem-forge/actions/workflows/ci.yml/badge.svg"></a>
  <img alt="License Apache 2.0" src="https://img.shields.io/badge/license-Apache--2.0-blue">
  <img alt="Status beta" src="https://img.shields.io/badge/status-beta-3b82f6">
</p>

<p align="center">
  <picture>
    <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/dodo-labs-ai/mem-forge/main/.github/assets/demo-dark.gif">
    <img src="https://raw.githubusercontent.com/dodo-labs-ai/mem-forge/main/.github/assets/demo-light.gif" alt="A new hire asks a coding agent how to answer a support ticket. The agent searches MemForge and answers with two team rules, citing a GitHub Pages doc and a Jira ticket." width="100%">
  </picture>
</p>

> **Status: beta.** APIs, storage formats and plugin packaging may still change.
> A new release saves a recovery point of the database and vector index before
> it upgrades them, rolls back a migration that fails, and refuses data a newer
> release wrote; restoring the recovery point goes back
> ([Upgrading](https://github.com/dodo-labs-ai/mem-forge/blob/main/docs/quickstart.md#upgrading)).
> The open-source build runs for one user on one machine.

## Quick start

You need Docker with Compose v2 and an API key for a model provider
(OpenRouter, OpenAI or Anthropic), or a model server such as Ollama.

```bash
git clone https://github.com/dodo-labs-ai/mem-forge.git && cd mem-forge
cp .env.example .env
docker compose up --build
```

Open the admin UI at http://localhost:5174 and choose a generation model and an
embedding model in Settings. Then install the agent plugin:

```text
# Claude Code (inside a session)
/plugin marketplace add dodo-labs-ai/mem-forge
/plugin install memory@memforge
```

```bash
# Codex
codex plugin marketplace add dodo-labs-ai/mem-forge
codex plugin add memory@memforge
```

Open **Sources** in the admin UI to connect Confluence, Jira or another source,
then start a new agent session and ask:

```text
Before you change this, check MemForge for decisions and conventions that
apply, and show the source for each one you rely on.
```

When you settle something in a session, ask the agent to remember it. It shows
you the memory first and saves it after you confirm.

The [quickstart](https://github.com/dodo-labs-ai/mem-forge/blob/main/docs/quickstart.md) covers model settings,
[plugin routing](https://github.com/dodo-labs-ai/mem-forge/blob/main/docs/quickstart.md#plugin-routing) for a hosted endpoint or
more than one workspace, local folders, the host CLI and network details.

## Why MemForge

**One memory for your docs and your decisions.** Confluence, Jira, Notion,
GitHub and Teams land in one store and come back from one `search` call. When
you tell Claude Code or Codex to remember a decision or a correction, it saves
that one memory after you confirm, never a transcript. A decision you saved in
Claude Code is there when you open Codex.

**It follows the source.** When a page or ticket changes, the next sync
rechecks the memories built on it. Each one is kept, replaced or retired, and
earlier versions keep their history. Across
documents, MemForge marks memories that repeat, update or contradict each
other.

<p align="center">
  <img src="https://raw.githubusercontent.com/dodo-labs-ai/mem-forge/main/.github/assets/memory-detail.png" alt="A memory in the admin UI: a decision from a Jira ticket that updates an older memory from the support playbook, shown with the ticket text it rests on." width="100%">
</p>

**Every answer has a source.** `get_memory` returns the supporting text and a
link back to the page, ticket or file. Where the source allows it, the citation
is pinned to the revision and passage that was read, and `get_resource` opens
the original. The supporting text can be the model's focused wording, not
always a verbatim quote.

**Runs on your machine.** Apache-2.0. `docker compose up` starts the service
with local storage (SQLite, FTS5 and Chroma). You bring your own model
endpoints and keys, and those providers receive what is sent for inference.
The plugins send no session transcripts: an agent writes only the memory you
confirm.

See [how MemForge compares](https://github.com/dodo-labs-ai/mem-forge/blob/main/docs/comparison.md) with Mem0, Zep, claude-mem and
Unblocked.

## How it works

```mermaid
flowchart LR
  subgraph You["Your machine"]
    CC["Claude Code / Codex"]
    Plugin["MemForge plugin<br/>MCP proxy"]
  end
  Docs["Confluence, Jira, Notion,<br/>GitHub, Teams, local Markdown"]
  API["MemForge API"]
  Extract["Extract memories<br/>with source Evidence"]
  Recheck["Plan updates<br/>relate memories"]
  Store[("SQLite + FTS5<br/>Chroma")]

  CC --> Plugin
  Docs -- "sync" --> API
  API --> Extract --> Recheck --> Store
  Plugin -- "search / get_memory / get_resource<br/>create_memory (confirmed)" --> API
  API -- "read" --> Store
```

1. **Capture.** Connectors sync documents, and an agent saves a memory you ask
   it to remember after you confirm it.
2. **Extract.** An LLM turns each source into short memories with supporting
   Evidence.
3. **Reconcile.** When a source changes, affected memories are rechecked and
   related ones are linked.
4. **Recall.** Agents call `search`, `get_memory` and `get_resource` over MCP.

More detail: [architecture overview](https://github.com/dodo-labs-ai/mem-forge/blob/main/docs/architecture.md).

## Integrations

MemForge sits between the systems where your team writes things down and the
agents that do the work.

**Agent integrations**

<p>
  <img src="https://raw.githubusercontent.com/dodo-labs-ai/mem-forge/main/.github/assets/icons/claude.svg" width="18" height="18" align="absmiddle" alt="">&nbsp;Claude&nbsp;Code&nbsp;&nbsp;&nbsp;
  <picture><source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/dodo-labs-ai/mem-forge/main/.github/assets/icons/openai-dark.svg"><img src="https://raw.githubusercontent.com/dodo-labs-ai/mem-forge/main/.github/assets/icons/openai-light.svg" width="18" height="18" align="absmiddle" alt=""></picture>&nbsp;Codex
</p>

**Memory sources**

<table>
  <tr>
    <td>Docs and wikis</td>
    <td nowrap>
      <img src="https://raw.githubusercontent.com/dodo-labs-ai/mem-forge/main/.github/assets/icons/confluence.svg" width="18" height="18" align="absmiddle" alt="">&nbsp;Confluence&nbsp;&nbsp;&nbsp;
      <picture><source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/dodo-labs-ai/mem-forge/main/.github/assets/icons/notion-dark.svg"><img src="https://raw.githubusercontent.com/dodo-labs-ai/mem-forge/main/.github/assets/icons/notion-light.svg" width="18" height="18" align="absmiddle" alt=""></picture>&nbsp;Notion&nbsp;&nbsp;&nbsp;
      <picture><source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/dodo-labs-ai/mem-forge/main/.github/assets/icons/github-dark.svg"><img src="https://raw.githubusercontent.com/dodo-labs-ai/mem-forge/main/.github/assets/icons/github-light.svg" width="18" height="18" align="absmiddle" alt=""></picture>&nbsp;GitHub&nbsp;Pages&nbsp;&nbsp;&nbsp;
      <picture><source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/dodo-labs-ai/mem-forge/main/.github/assets/icons/github-dark.svg"><img src="https://raw.githubusercontent.com/dodo-labs-ai/mem-forge/main/.github/assets/icons/github-light.svg" width="18" height="18" align="absmiddle" alt=""></picture>&nbsp;GitHub&nbsp;repo&nbsp;docs&nbsp;&nbsp;&nbsp;
      <img src="https://raw.githubusercontent.com/dodo-labs-ai/mem-forge/main/.github/assets/icons/folder.svg" width="18" height="18" align="absmiddle" alt="">&nbsp;Local&nbsp;folders
    </td>
  </tr>
  <tr>
    <td>Tickets</td>
    <td nowrap><img src="https://raw.githubusercontent.com/dodo-labs-ai/mem-forge/main/.github/assets/icons/jira.svg" width="18" height="18" align="absmiddle" alt="">&nbsp;Jira</td>
  </tr>
  <tr>
    <td>Chat</td>
    <td nowrap><img src="https://raw.githubusercontent.com/dodo-labs-ai/mem-forge/main/.github/assets/icons/teams.svg" width="18" height="18" align="absmiddle" alt="">&nbsp;Teams</td>
  </tr>
  <tr>
    <td>Planned</td>
    <td nowrap>
      <img src="https://raw.githubusercontent.com/dodo-labs-ai/mem-forge/main/.github/assets/icons/obsidian.svg" width="18" height="18" align="absmiddle" alt="">&nbsp;Obsidian
    </td>
  </tr>
</table>

[Add a source](https://github.com/dodo-labs-ai/mem-forge/blob/main/docs/quickstart.md#4-add-a-source) lists what each one syncs and
how to scope it. Need one that is missing?
[Open an issue](https://github.com/dodo-labs-ai/mem-forge/issues).

## Limits

- The open-source build is single-user, has no authentication, and binds to
  `127.0.0.1` only. Do not expose it on a network.
- Model extraction and semantic judgments can miss useful knowledge or make
  mistakes. Check the cited Evidence before relying on a consequential claim.
- A generation model is required; without one, syncs store content but create
  no memories.

## Documentation

- [Quickstart](https://github.com/dodo-labs-ai/mem-forge/blob/main/docs/quickstart.md): setup, models, plugins, sources, CLI,
  development from source
- [Docs index](https://github.com/dodo-labs-ai/mem-forge/blob/main/docs/README.md): quickstart, architecture, API and release guides
- [Contributing](https://github.com/dodo-labs-ai/mem-forge/blob/main/CONTRIBUTING.md)

## License

Apache License 2.0. See [LICENSE](https://github.com/dodo-labs-ai/mem-forge/blob/main/LICENSE).