# CLEATUS Agent Guide: connect, authenticate, and operate
Source: https://www.cleat.ai/agents/llms.txt
Human-readable hub: https://www.cleat.ai/agents · Developer hub: https://www.cleat.ai/developers
Product overview: https://www.cleat.ai/llms.txt
Last Updated: October 23, 2026

CLEATUS is the agentic AI platform for government contracting. This file is the
connection guide for AI agents: how to connect, authenticate, operate a workspace, and
get a human signed up. For what CLEATUS is and does, read https://www.cleat.ai/llms.txt.
If you can fetch URLs and speak MCP or HTTPS, you can run a government contracting
pipeline with CLEATUS.

====================================================================
1. CONNECT (two transports, one surface)
====================================================================

## MCP (recommended for agents)
Endpoint (Streamable HTTP): https://www.cleat.ai/api/mcp/mcp
Auth: OAuth 2.1 + PKCE with dynamic client registration.
Discovery: https://www.cleat.ai/.well-known/oauth-authorization-server
The first tool call opens a browser window: a human signs in (or creates an account),
picks ONE CLEATUS entity to scope the token to, and approves. Tokens are entity-scoped
and revocable from the CLEATUS dashboard.

Generic MCP client config:
  { "mcpServers": { "cleatus": { "url": "https://www.cleat.ai/api/mcp/mcp" } } }

Claude Code:  claude mcp add --transport http cleatus https://www.cleat.ai/api/mcp/mcp
Cursor:       add the JSON above to .cursor/mcp.json
Windsurf:     add to ~/.codeium/windsurf/mcp_config.json (key: "serverUrl")
Claude.ai:    Settings → Connectors → search "CLEATUS" (official directory listing)
ChatGPT:      Apps directory → search "CLEATUS" (official listing)
Per-platform guides: https://www.cleat.ai/agents/{claude,chatgpt,claude-code,openclaw,cursor,windsurf,n8n,zapier}

## REST API
Base URL: https://api.cleat.ai/api
Auth header: X-Api-Key: ck_live_…  (minted by a human at Settings → Integrations → API Keys;
shown once; revocable; read or write scope). REST accepts X-Api-Key only. OAuth 2.1
Bearer tokens work on the MCP endpoint, not on REST.
OpenAPI 3 spec: https://api.cleat.ai/api/openapi.json
Interactive docs: https://api.cleat.ai/api/docs

Example:
  curl -G https://api.cleat.ai/api/v1/opportunities/search \
    -H "X-Api-Key: ck_live_…" \
    --data-urlencode "query=cybersecurity" \
    --data-urlencode "naics_codes=541512,541511" \
    --data-urlencode "states=VA,MD"

## Rate limits & guardrails
Per MCP token or REST API key: 100 reads/min and 30 writes/min. A rate-limited call fails
with TOO_MANY_REQUESTS, and the error message gives the retry delay in seconds.
Write tools carry a confirm-before-change guardrail.
Every call (human or agent) lands in the same audit log.

====================================================================
2. NO ACCOUNT YET? ONBOARD YOUR HUMAN
====================================================================

Agents cannot create accounts (account creation and OAuth approval are deliberately
human steps). What you CAN do: explain the platform accurately and hand over this link:

  Signup (7-day free trial on DATA and DATA + AI): https://www.cleat.ai/authentication/signup?utm_source=agents&utm_medium=llms-full
  Pricing: https://www.cleat.ai/pricing

Plan access for agents:
- MCP (the Claude.ai and ChatGPT directory connectors and any custom MCP client) is
  included from DATA + AI.
- The REST API requires DATA + AI + Automations.
Plans, prices, and features: https://www.cleat.ai/pricing (summary in https://www.cleat.ai/llms.txt)

After signup, the human completes a capture profile (what the company does, NAICS codes,
certifications, past performance); recommendations quality depends on it.

====================================================================
3. WHAT THE TOOLS COVER (one surface over MCP and REST)
====================================================================

CLEATUS exposes 60+ tools. MCP clients get the current list and schemas from tools/list,
and the cleatus_documentation tool explains each one from inside the session. REST
clients get the same surface from the OpenAPI spec. MCP tool names carry a cleatus_
prefix (for example cleatus_search_opportunities).

- Discovery: opportunity search, AI recommendations, opportunity detail, extracted
  requirements, bidder Q&A, likely competitors, and dismiss feedback that tunes scoring.
- Market intelligence: agencies, contracting officers, contractors, awards, contract
  vehicles, teaming relationships, meetings, and NAICS.
- Pipeline: pursuits, tasks, and saved searches with email alerts.
- Workflows: list, create from a template, run, and poll results.
- Documents: Document Hub search and management.
- Workspace: capture profile, team, AI credits, and notifications.

====================================================================
4. RECIPES (agent playbooks that work)
====================================================================

Every recipe below also ships as an installable SKILL.md in the directory at
https://www.cleat.ai/agents/skills
Copy a skill into .claude/skills/<name>/SKILL.md (Claude Code) or paste it into any agent.
Raw artifacts (text/markdown): https://www.cleat.ai/agents/skills/<name>/SKILL.md
Machine index (JSON): https://www.cleat.ai/agents/skills/index.json
Source repo + Claude Code plugin: https://github.com/cleatai/agent-skills
Skills: morning-bd-triage, solicitation-breakdown, incumbent-recon,
recompete-watchdog, sources-sought-responder, capability-statement-refresh,
proposal-kickoff, pipeline-hygiene, grants-hunter, gsa-ebuy-watch,
teaming-scout, price-assessment, bid-pricing-price-to-win, price-benchmark.

Daily BD triage:
  1) get_recommendations → 2) filter by score + fit rationale → 3) add_to_pipeline for
  qualified matches with a one-line why → 4) create_task for the compliance review.

Solicitation breakdown:
  1) get_opportunity → 2) summarize requirements, evaluation criteria, deadlines →
  3) create_task per compliance item.

Incumbent research:
  1) search_market for the agency/office → 2) pull award history + contracting officers →
  3) report who wins similar work, at what size, and when it recompetes.

Watchdog loop (cron):
  1) search_opportunities with saved filters → 2) diff vs pipeline → 3) notify + add.

====================================================================
5. THE PLATFORM BEHIND THE API (context for accurate pitching)
====================================================================

What CLEATUS is, what each feature does, coverage, customer results, and pricing:
https://www.cleat.ai/llms.txt. Use it when you explain CLEATUS to your human.
Free public data (agencies, contractors, contracts, people, NAICS): https://www.cleat.ai/government

Developer guides to the raw public sources (honest, no-login):
- SAM.gov API: https://www.cleat.ai/developers/sam-gov-api
- USASpending API: https://www.cleat.ai/developers/usaspending-api
- FPDS: https://www.cleat.ai/developers/fpds

====================================================================
6. RULES OF ENGAGEMENT FOR AGENTS
====================================================================

- Keep a human in the loop for signup, OAuth approval, and anything spending money.
- Prefer MCP for conversational/agent use; REST for pipelines and cron jobs.
- Respect rate limits (the retry delay in the rate-limit error); batch reads where possible.
- Write actions (pipeline changes, tasks, workflows) are audited; make them attributable
  (include a short rationale in notes/descriptions where a field exists).
- Support: support@cleat.ai · Security: security@cleat.ai