Quickstart¶
5-minute path¶
1. Clone and install¶
git clone https://github.com/navendubrajesh/context-management-for-agents.git
cd context-management-for-agents
pip install -e runtime/core -e runtime/mcp -e runtime/api
2. Run the proof demo¶
python examples/demo/run_demo.py
python examples/demo/run_mcp_demo.py
Expected: router picks context-compression; handoff compaction shows ~93% token reduction on the bundled conversation fixture.
3. Register MCP in Cursor / Copilot¶
.cursor/mcp.json (adjust paths):
{
"mcpServers": {
"context-skills": {
"command": "python",
"args": ["server.py"],
"cwd": "/path/to/context-management-for-agents/runtime/mcp",
"env": {
"PYTHONPATH": "/path/to/context-management-for-agents/runtime/core"
}
}
}
}
Tools: list_skills, route_task, get_skill, compact_session, run_context_pipeline, and more.
4. One-line REST smoke test¶
uvicorn app:app --app-dir runtime/api --port 8080 &
curl -s http://localhost:8080/skills | head -c 200
5. Install skills into your editor (optional)¶
npx context-management-for-agents --platform cursor
npm package¶
npm i context-management-for-agents
See the root README for all platform targets.
Next steps¶
- Runtime reference
- Benchmarks — reproduce measurements
- Enterprise deployment — SSO, tenancy, policy (Phase 3)