rastro.ai 0.2.0

MCP server for Rastro catalog operations — 26 tools for schema, items, activities, snapshots, AI mapping, judging, and image editing

rastro-mcp

Standalone MCP server for Rastro catalog workflows.

What this is

• API-key-only MCP server for Codex/Claude
• No direct database access needed
• Activity-first safety model (stage -> review in dashboard -> apply)

Install

git clone https://github.com/Rastro-AI/rastro-mcp.git
cd rastro-mcp
uv sync

Required environment

# Option A (recommended): API key auth
export RASTRO_API_KEY="rastro_pk_..."

# Option B: bearer token from web auth session
# export RASTRO_ACCESS_TOKEN="<jwt>"

export RASTRO_BASE_URL="https://catalogapi.rastro.ai/api"
# optional
# export RASTRO_ORGANIZATION_ID="<org_uuid>"

Browser login from CLI (no copy/paste)

uv run rastro-mcp login
# then apply the printed export in your shell:
# export RASTRO_ACCESS_TOKEN='...'

This opens dashboard.rastro.ai, uses your existing web session, and redirects back to a localhost callback.

Run (stdio)

uv run rastro-mcp

Equivalent:

uv run python -m rastro_mcp.server

Codex / Claude MCP config

Use a workspace .mcp.json entry like:

{
  "mcpServers": {
    "rastro": {
      "command": "bash",
      "args": [
        "-lc",
        "cd /ABSOLUTE/PATH/rastro-mcp && export RASTRO_API_KEY='rastro_pk_...' && export RASTRO_BASE_URL='https://catalogapi.rastro.ai/api' && uv run rastro-mcp"
      ]
    }
  }
}

Safety defaults

• Programmatic approve/apply is disabled.
• Use dashboard review URL to approve/apply.
• Large staged updates are chunked internally into one activity.

Docs

• Quickstart: https://docs.rastro.ai/mcp/quickstart
• Reference: https://docs.rastro.ai/mcp/reference

License

MIT

PyPI Publish Notes

• GitHub Actions workflow: .github/workflows/publish.yml
• Trusted publisher values on PyPI must match:
  • Owner: Rastro-AI
  • Repository: rastro-mcp
  • Workflow: publish.yml