Visible multi-agent orchestrator and MCP server for Kimi Code K3. It controls Kimi through ACP (JSON-RPC over stdio), rather than attempting to drive its terminal UI.

Start

dotnet run --project KimiFleet.csproj -- --port 7373

Open a second terminal and watch live activity:

curl -N http://127.0.0.1:7373/events

Open http://127.0.0.1:7373 for the dashboard: it shows agents, their scoped ownership, state and a human-readable live activity feed.

MCP mode

Start both the dashboard and an MCP stdio server:

dotnet run --project KimiFleet.csproj -- --mcp --port 7373

The server implements fleet_list_agents, fleet_start_agent, fleet_assign_task, fleet_request_review, and fleet_stop_agent. In MCP mode stdout is protocol-only; operational logs are written to stderr and remain visible in the dashboard.

API

Create an agent with exclusive source scopes:

curl -X POST http://127.0.0.1:7373/agents \
  -H 'Content-Type: application/json' \
  -d '{"name":"vulkan","workspace":"/home/emil/Desktop/Cortex_Engine","role":"renderer","scopes":["src/Engine.Graphics.Vulkan","src/Engine.Graphics"]}'

Give it work:

curl -X POST http://127.0.0.1:7373/agents/vulkan/prompt \
  -H 'Content-Type: application/json' \
  -d '{"prompt":"Read AGENTS.md, inspect the Vulkan backend, and propose one safe improvement."}'

Ask another agent to review the current diff without editing:

curl -X POST http://127.0.0.1:7373/agents/reviewer/review \
  -H 'Content-Type: application/json' \
  -d '{"author":"vulkan","context":"Review the Vulkan change after its tests finish."}'

Useful endpoints: GET /health, GET /agents, GET /events, POST /agents/{name}/stop.

All ACP children launch as kimi --yolo --model kimi-code/k3 acp. Scope locks prevent two fleet agents from being assigned the same area, but they do not replace a final git status check before edits.