Integrations
Every assistant. One connection.
TenderGraph is built agent-first: the same MCP server that powers our own platform plugs into Claude, ChatGPT, Perplexity and Grok — and the REST API, OpenAPI spec and webhooks cover everything else. Two minutes from API key to your agent answering “what should we bid on this week?”
How the agent connection works
- 1
Create your API key
Generated during onboarding, managed any time under API keys. Keys are scoped, revocable, and rate-limited.
- 2
Paste one endpoint
Add the MCP endpoint to your assistant's connector settings — exact steps for each platform below.
- 3
Ask real questions
Your agent gets fifteen tools: it searches, qualifies against your profile, analyses winners, and watches your market between sessions.
AI assistants
Claude
claude.ai · Claude Desktop · Claude CodeSettings → Connectors → Add custom connector
Add the endpoint and sign in once in the browser where OAuth is offered — or paste the endpoint with your key on it. Claude picks up all fifteen tools plus guided prompts for finding and qualifying tenders.
ChatGPT
Plus · Pro · Business (Developer mode)Settings → Apps & Connectors → Developer mode → Create
Create a connector with the endpoint and OAuth where offered, or the key-in-path URL with no authentication. The search/fetch pair follows OpenAI's deep-research contract, so TenderGraph also works as a ChatGPT research source; API builders attach it as a remote MCP tool in the Responses API.
Perplexity
Pro · Max · EnterpriseAccount Settings → Connectors → + Custom connector
Add the endpoint as a remote connector with API-key auth, then enable it as a source in any thread — Perplexity cites open tenders in its answers.
Grok
grok.com · xAI APIgrok.com/connectors → New Connector → Custom
Point Grok at the endpoint and authenticate with your key. On the xAI API, pass the server as a remote MCP tool with your key as the authorization value.
Exact setup commands for each platform: connection guides.
For builders
REST API
Any language, any stackAuthorization: Bearer tg_live_…
Search, full records with extracted document text where available (open tenders first), qualification scoring, award analytics, monitors — JSON responses designed to be read by models as much as by code.
OpenAPI 3.1
Codegen & custom GPTsGET /api/v1/openapi
The complete typed spec — generate clients, wire up a custom GPT Action, or hand it to your platform team.
Webhooks
Push, not pollHMAC-signed POSTs to your endpoint
Attach a webhook to any monitor and new matching tenders arrive at your systems on the next delivery run after we first see them — signed, ordered, and delivered with retries and backoff.
MCP (any client)
The open agent standardStreamable HTTP at /api/mcp
Anything that speaks Model Context Protocol connects the same way — including stdio-only clients via the mcp-remote bridge.
What your agent can do
search_tenders- filtered search over every open and historical notice
get_tender- the full record: documents, contacts, provenance
search_companies- find a supplier — yours, a competitor, an incumbent
get_company- their full award history: value, buyers, sectors, trend
qualify_tender- bid/no-bid scoring against your company profile
analyze_awards- who wins, from which buyers, at what value
create_monitor / check_monitor- “what's new since I last looked”
save_profile / list_profiles- your targeting, stored and reusable
list_sources / get_stats- coverage, licences and each source's last sync, machine-readable
search / fetch- the deep-research pair for research agents
Write actions (saving profiles, creating monitors) require the key's write scope; everything is rate-limited per key and revocable instantly.
Before you connect anything
Try it with your own agent
Every trial includes full MCP and API access from minute one — your key is created during onboarding.
Start 7-day free trial