Without a coordination layer, multi-agent setups break in predictable ways.
Two agents start dev servers on the same port.
Agent A discovers something Agent B needs, no sharing mechanism.
Agents crash without cleanup.
Everything your agents need to work together without stepping on each other.
Atomic port, file, and service claiming with TTL and conflict detection.
Vector-indexed shared memory with RAG search across all agent discoveries.
SSE event stream for instant coordination and command delivery.
Agents post screenshots and test results, humans see health dashboards.
AES-256-GCM encrypted secrets and config shared across agents.
Native MCP tools for Claude Code, Cursor, and any MCP-compatible agent.
Claude Code
Cursor
Windsurf
Custom Agents
via REST API or CLI
Install the CLI, run the setup wizard, and your agents coordinate automatically. Works with Claude Code, Cursor, or any tool that can run shell commands.
# Install globally (or use npx)
npm install -g tm-agent
# Interactive setup wizard
# Creates API key, configures your tool
tm-agent setup
# Writes .env.agent + MCP config for
# Claude Code, Cursor, or both# Start local daemon (background)
tm-agent daemon --name my-machine &
# Agents use MCP tools natively:
# tm_claim_resource port:3000
# tm_add_knowledge "auth uses JWT"
# tm_search_knowledge "how auth works"
# Or use the CLI directly:
tm-agent claim port:3000 --ttl 3600
tm-agent peers