timps-swarm CLI

160 AI specialists — code, security, DevOps, AI/ML, infra, India-stack, health, research, content, voice, robotics, and more. One npm install puts every one of them into your AI coding tool as parallel sub-agents. Works with just Node — no Python required.

Install

npm install -g timps-swarm
# or run without installing:
npx timps-swarm <command>

The postinstall hook auto-detects every AI tool on your machine and:

1. Writes the timps-swarm MCP server entry into each IDE config file, including an explicit env: block that forwards your API keys (GEMINI_API_KEY, ANTHROPIC_API_KEY, OPENAI_API_KEY, GROQ_API_KEY, TIMPS_API_URL, OLLAMA_HOST, REDIS_URL).
2. Writes one sub-agent .md file per agent (160 total) into ~/.claude/agents/, ./.claude/agents/, and ~/.codex/agents/ so Claude Code, Cursor, and Codex can dispatch them in parallel via Task(subagent_type="timps_kubernetes_navigator").

Restart your tool — 160 agents appear as MCP tools and as parallel sub-agents.

# Zero-setup, no install required:
npx timps-swarm audit ./          # security scan any repo

Run without the Python backend

npm install -g timps-swarm works without the Python repo. The CLI ships cli/lib/mcp-proxy.js, a Node.js JSON-RPC 2.0 stdio proxy that forwards every MCP tool call to a running FastAPI server (local or remote via TIMPS_API_URL). The full Python server is only needed for MCP sampling (using the host's LLM) and for tools that have to run locally; everything else works through the proxy.

# Point at a remote API
TIMPS_API_URL=https://timps.example.com npx timps-swarm mcp

# Run the Python server yourself (only if you want the full experience)
git clone https://github.com/Sandeeprdy1729/timps-swarm
cd timps-swarm
pip install -e ".[dev]"
make up-local                       # starts FastAPI on :8000

Commands

audit — security scan (works offline)

npx timps-swarm audit ./                       # scan CWD
npx timps-swarm audit ./src --ecosystem python # Python focus
npx timps-swarm audit requirements.txt
npx timps-swarm audit package.json --ecosystem node

fix — run the full 10-agent SDLC pipeline

npx timps-swarm fix "build a REST API for a todo app"
npx timps-swarm fix ./src --language typescript

health — diagnose your machine (12 specialist agents)

npx timps-swarm health

Research, design, scaffolding

npx timps-swarm research "best practices for distributed tracing in 2026" --depth deep
npx timps-swarm api-design "billing API with metered usage and Stripe webhooks"
npx timps-swarm db-design "multi-tenant SaaS with usage-based billing" --engine postgresql
npx timps-swarm n8n "when GitHub issue is labeled bug, post to Slack and create Jira ticket"

providers — show LLM provider status

npx timps-swarm providers

mcp — start the MCP stdio server

npx timps-swarm mcp                            # auto-detect Python repo
npx timps-swarm mcp --repo /path/to/repo       # explicit
TIMPS_API_URL=https://... npx timps-swarm mcp  # use remote API

For use in ~/.claude/mcp.json:

{
  "mcpServers": {
    "timps-swarm": {
      "command": "npx",
      "args": ["timps-swarm", "mcp"],
      "env": {
        "GEMINI_API_KEY": "...",
        "ANTHROPIC_API_KEY": "..."
      }
    }
  }
}

start — start the FastAPI server (Python repo required)

npx timps-swarm start                # start in foreground on :8000
npx timps-swarm start --repo ./..    # start with explicit repo path

install-mcp / uninstall-mcp — wire up (or remove) every IDE

npx timps-swarm install-mcp              # all detected tools + 160 sub-agents
npx timps-swarm install-mcp --tool cursor          # one tool only
npx timps-swarm install-mcp --no-sub-agents        # MCP config only, no .md files
npx timps-swarm install-mcp --dry-run              # preview without writing
npx timps-swarm uninstall-mcp           # remove the MCP config and 160 sub-agents

By default install-mcp writes:

• MCP server entries into 9 IDE config files (one per IDE, all pointing at npx timps-swarm mcp).
• 160 sub-agent .md files into ~/.claude/agents/, ./.claude/agents/, ~/.codex/agents/ (one per MCP tool) so Claude Code's Task(subagent_type="timps_kubernetes_navigator"), Cursor Composer, and Codex can dispatch them in parallel.

All writes are idempotent (re-running updates the existing file) and reversible via uninstall-mcp (only removes the timps-swarm key and the timps-*.md files — your other config is untouched).

Flags

FLAGS (install-mcp)
  --tool <id>            Configure one IDE only
                         (claude-code, cursor, codex-cli, windsurf, vscode, …)
  --no-sub-agents        Skip writing 160 sub-agent .md files (MCP config only)
  --dry-run              Preview without writing
  --silent               Suppress output (postinstall)

FLAGS (start / mcp)
  --repo <path>          Explicit path to the Python repo (skip auto-detection)

Environment Variables

| Variable | Description | | :--- | :--- | | TIMPS_API_URL | TIMPS server URL — default http://localhost:8000. The Node.js MCP proxy, install-mcp, and providers all honour this. | | TIMPS_REPO | Explicit path to the Python repo. Skips findRepoPython() auto-detection (priority: $TIMPS_REPO → walk up CWD → ~/timps-swarm → ~/Desktop/timps-swarm → /opt/timps-swarm → npm global prefix). | | GEMINI_API_KEY | Google Gemini — free tier, recommended. | | ANTHROPIC_API_KEY | Anthropic Claude. | | OPENAI_API_KEY | OpenAI GPT-4o. | | GROQ_API_KEY | Groq (fast Llama 3.3 70B). | | OLLAMA_HOST | Ollama server URL — default http://localhost:11434. | | REDIS_URL | Redis server URL — default redis://localhost:6379/0. Falls back to in-memory dict silently when Redis is down. | | LOG_LEVEL | Default INFO. |

The 160 agents

9 categories. Every one is callable as an MCP tool and registered as a native Claude Code / Cursor / Codex sub-agent.

| Category | Count | Highlights | |----------|------:|------------| | Priority | 68 | research_agent, ab_testing_agent, abdm_agent, browser_automation, churn_predictor, deep_research_agent, digilocker_agent, dpdp_act_auditor, federated_learning, finetuning_agent, fssai_compliance_agent, gst_compliance, indiehacker_agent, model_evaluator, model_perf_monitor, prompt_injection_scanner, quantum_ready, rag_designer, rag_evaluator, red_team_agent, sbom_generator, service_mesh_configurator, storybook_story_generator, upi_agent, voice_agent_designer, wearable_health_coach, web3_agent, win_loss_analyst, … | | Expert Diagnostics | 51 | kubernetes_navigator, docker_compose_architect, pipeline_healer, mcp_server_generator, observability_cost_optimizer, accessibility_tester, license_compliance_scanner, iac_drift_detector, disaster_recovery, contract_reviewer, court_case_summarizer, data_pipeline, db_migration_pilot, game_day_facilitator, git_workflow_automator, graphql_agent, load_testing, local_rag_builder, log_pattern_analyzer, phishing_simulator, postmortem_agent, test_intelligence, visual_regression_detective, web_scraping, web_search, … | | Computer Health | 12 | system_optimizer, file_organizer, environment_doctor, security_guard, network_medic, battery_analyst, update_manager, log_interpreter, privacy_cleaner, media_librarian, backup_sentinel, context_switcher | | Developer Workflow | 12 | issue_triager, boilerplate_architect, pr_reviewer, dependency_sentinel, unit_test_writer, docstring_generator, log_detective, sql_optimizer, sprint_reporter, flaky_test_hunter, api_contract_auditor, content_multiplier | | Knowledge Worker | 7 | inbox_gatekeeper, meeting_condenser, research_scout, trend_monitor, data_wrangler, competitor_tracker, agri_commodity_forecaster | | Meta | 6 | list_agents, dispatch, full_checkup, list_providers, connect_tools, tool_status | | Context / Kernel | 3 | context_briefing, delegate, kernel_status | | SDLC Pipeline | 1 | run_task — the 10-node LangGraph orchestrator: PM → Architect → Code → Review → QA → Security → Perf → Docs → DevOps |

# Trigger the full SDLC pipeline
python3 give_work.py "Build a rate-limited REST API for user authentication"

# Run a computer health checkup
npx timps-swarm health
python3 give_work.py "My laptop fan is always running"

# Delegate a multi-step goal
npx timps-swarm delegate "fix the auth bug and ensure 80% test coverage"

Requirements

• Node.js 18+
• One of: a running FastAPI server (Python repo OR TIMPS_API_URL set), or just the Node.js fallback proxy (calls any reachable API).

License

MIT