Integrations

Aider

Persistent cross-session memory for Aider. Decisions from last week's session are available in today's — without re-pasting context.

Why Waystone + Aider

Aider is excellent at coding against your repository, but each session starts cold. Architectural decisions, rejected approaches, and implementation constraints live in your head, not Aider's. Waystone fixes this:

• Retrieves only what's relevant to the current file or task — not a flat dump of everything
• Persists rejected approaches so Aider doesn't re-suggest them the next session
• Accumulates project-specific decisions (naming conventions, API choices, architecture) across months of work
• Runs locally — no data leaves your machine unless you opt into the hosted API

1. Install Waystone

# CLI in early access — request at unbidden.ai

Verify the install:

waystone --version

2. Add to your Aider MCP config

Create or edit ~/.aider/mcp.json (global) or .aider/mcp.json in your project root:

{
  "mcpServers": {
    "waystone": {
      "command": "waystone",
      "args": ["mcp-serve"],
      "env": {
        "WAYSTONE_PROJECT": "my-project"
      }
    }
  }
}

Replace my-project with the name for this project's memory store. Then launch Aider pointing at the config:

aider --mcp-config ~/.aider/mcp.json

Or add it to your .aider.conf.yml so it applies to every session in the project:

mcp-config: .aider/mcp.json

Once connected, you'll see waystone_query, waystone_extract, waystone_stats, and waystone_list_projects in the available tools.

3. Set your system prompt

Tell Aider to use Waystone memory at session start and end. Add this to your --system-prompt or create a .aider/system_prompt.md:

## Memory (Waystone)
At the start of each session, call waystone_query with a brief description of
the current task or the file you're about to edit. This retrieves relevant
decisions, constraints, and architectural context from prior sessions.

After completing significant work — a feature, a refactor, a tricky bug fix —
call waystone_extract with a summary of what was changed and why. Focus on
decisions that would save time if you had to re-explain them next session.

Do not duplicate memory into comments or NOTES.md. Use waystone_extract instead.

4. Initialize the project

Before the first session, initialize the Waystone project:

waystone init my-project

If you have an existing NOTES.md or architecture document, seed it now:

waystone extract my-project NOTES.md

Verify the import:

waystone show my-project

How it works in practice

Once configured, Aider sessions follow this pattern automatically:

1. Session start: Aider calls waystone_query with the current task. Waystone returns the 10–25 most relevant facts — naming conventions, prior decisions about the file being edited, rejected approaches.
2. During the session: You edit files, Aider applies changes. waystone_query may be called again as the scope evolves.
3. Session end: Aider calls waystone_extract with a summary of what changed and why. Waystone extracts structured facts, detects which supersede prior decisions, and updates the graph.

Hosted API (optional)

By default, Waystone runs locally — memory is a SQLite file in ~/.waystone/. For cross-machine access:

{
  "mcpServers": {
    "waystone": {
      "command": "waystone",
      "args": ["mcp-serve"],
      "env": {
        "WAYSTONE_PROJECT": "my-project",
        "WAYSTONE_API_KEY": "your-api-key",
        "WAYSTONE_REMOTE_URL": "https://api.waystone.unbidden.ai"
      }
    }
  }
}

Get an API key from the Pro or Team plan. Local mode is free and has no rate limits.

Troubleshooting