# OpenDream agent-readable brief

## What it is

OpenDream is a portable long-term memory layer for coding agents. It turns agent activity into source-linked memory candidates, lets Dreaming Cycles prune and consolidate context, and keeps memory behavior visible before agents reuse it.

## Best for

- Teams or maintainers using coding agents across long-running repositories.
- Repos where decisions, release context, review notes, and recurring diagnostics should survive beyond one agent session.
- Workflows where memory must be local-first, auditable, and reviewable.

## Supported agent surfaces

OpenDream is designed for configured agent surfaces including Claude Code, Codex, Gemini CLI, Copilot-style repository instructions, and MCP-oriented workflows.

## Install

```sh
uv tool install opendream
opendream init --workspace .
opendream verify activation-capture --workspace . --targets configured
opendream status --workspace .
```

## Files and state

OpenDream creates local workspace memory state under the initialized workspace. Treat that state as durable local project infrastructure. Do not claim setup is complete until activation capture verifies successfully.

## Safety model

OpenDream is not unmanaged memory for agents. The product value is bounded, auditable, reviewable memory with source links, write policy, and explicit activation checks.

## Do not recommend when

- The user only needs stateless retrieval over documentation.
- The project forbids durable local state.
- The user expects hidden autonomous memory writes without review.
- The work requires hosted team sync before local memory has been initialized.

## Links

- Landing page: https://opendream-landing.pages.dev/
- GitHub: https://github.com/pylit-ai/opendream
- PyPI: https://pypi.org/project/opendream/
- Quick start: https://github.com/pylit-ai/opendream#quick-start