DOCUMENTATION
QuickstartAuthenticationConcept MapTools ReferenceResources PromptsREST API MirrorRelated ReadingFAQMCP tools reference →REST API →
01
QuickstartHubbleField docs explain how the public web, REST, and MCP surfaces fit together, how clients authenticate, and which routes or tools to use for each discovery workflow. Use this page first if you need the integration contract rather than the marketplace browse experience.
1.For ChatGPT apps, point the client at https://hubblefield.ai/.well-known/mcp.json and complete the OAuth prompt.
2.For Claude Desktop or local MCP hosts, sign in at https://hubblefield.ai/signin and mint a bearer token.
3.Use the manual config below only for local stdio hosts.
claude_desktop_config.jsonJSON
{
  "mcpServers": {
    "deallayer": {
      "command": "npx",
      "args": ["-y", "@hubblefield/mcp"],
      "env": {
        "HUBBLEFIELD_TOKEN": "hf_live_••••••••"
      }
    }
  }
}
Remote clients should prefer the streamable-http endpoint and OAuth discovery instead of embedding a long-lived token.
02
AuthenticationStreamable-http clients authenticate with OAuth 2.1 authorization-code + PKCE. Manual bearer tokens remain available for local stdio MCP hosts and direct REST use.
MARKETPLACE OAUTH
Recommended for ChatGPT and other remote MCP clients. Uses dynamic client registration, PKCE, and revocation at /api/oauth/revoke.
MANUAL BEARER TOKEN
Fallback for local stdio hosts like Claude Desktop. Mint separately and pass via HUBBLEFIELD_TOKEN.
Free tier rate limits60 search calls / hour · 1,000 calls / day. Sufficient for human-driven agent workflows. Higher limits bundled with premium listings.
03
Concept MapUse these pages together: the glossary explains the operating model, the marketplace shows the public product, and the MCP reference defines the exact contracts that AI clients can rely on.
MarketplaceSee the public listing surface that the docs and glossary keep referring to.MCP tools referenceOpen the exact tool, resource, prompt, and REST mirror contracts.Glossary hubRead the short conceptual explanations behind the product model.Terms boundary pageUse the launch-phase terms stub for the clearest statement of public-use limits.
What is agent-first private-market discovery?Agent-first private-market discovery means company data is published in a structure that AI agents can search, compare, and cite directly instead of scraping narrative decks or cold outreach.How MCP fits private-market discoveryMCP gives AI clients a standard way to call tools such as company search and profile retrieval, which is why it matters for structured private-market discovery platforms.Public vs gated listing dataPublic listing data is the factual, indexable surface that can be safely browsed and cited, while gated data covers richer investor or diligence material that should not appear in search or public schemas.
04
Tools ReferenceCore tools available via MCP. The canonical request and response schemas live in MCP-TOOLS.md and mirror the REST endpoints under /api/v1.
search_companiesreadFull-text + filter search across published listings
get_companyreadRetrieve a single company profile by slug
list_raising_nowreadCompanies with an open raise, sorted by momentum
watchaccountAdd a company to your watchlist
request_introductionaccountSend a one-time intro request to the founder
upsert_listingwriteCreate or update a listing draft
add_metricswriteAppend a metrics period to a listing
add_monitoring_targetwriteRegister a URL for ongoing monitoring
whoamireadConfirm token identity and scopes
`read` tools work with marketplace OAuth or manual tokens. `account` tools act on the signed-in user. `write` tools require founder-side write scopes and are intended for draft management, metrics ingestion, and review workflows.
05
Resources & promptsThe MCP server also exposes reference resources and named prompts so a fresh agent can learn the workflow before it calls tools. Use resources/list and resources/read for product-copy guidance, and prompts/list / prompts/get for cookbook-style multi-step flows.
Available resourceshubblefield://guide/listing-rubric.mdhubblefield://guide/investor-playbook.mdhubblefield://guide/publishing-workflow.mdhubblefield://taxonomy/categories.mdhubblefield://policy/regulatory.mdhubblefield://schema/listing.jsonhubblefield://example/listing/<slug>.json
Available promptspublish_company_profileinvestor_briefmonitor_portfolio
06
REST API MirrorEvery MCP tool has a REST equivalent at /api/v1/<tool_name>. Same zod schemas, same response shape. Bearer token authentication.
POST/api/v1/search_companiesJSON body with SearchParams. Returns {results, count, _disclaimer}
GET/api/v1/get_company/:slugReturns {company, _disclaimer}. 404 if not found.
Search published companiesCURL
curl -X POST https://hubblefield.ai/api/v1/search_companies \
  -H "Authorization: Bearer hf_live_••••••••" \
  -H "Content-Type: application/json" \
  -d '{
    "q": "observability",
    "sectors": ["DevTools"],
    "raising_now": true,
    "limit": 10
  }'
Fetch one company by slugCURL
curl https://hubblefield.ai/api/v1/get_company/nimbus-ai \
  -H "Authorization: Bearer hf_live_••••••••"
Response envelopeAll responses include a _disclaimer field: “Information only. Not investment advice.” — so agent-rendered output picks it up automatically.
07
Related readingSupporting reference pages that explain the operating model behind the docs contracts.
Agent-first private-market discoveryExplains why the public listing surface is structured for search, citation, and AI retrieval.How MCP fits private-market discoveryGives the short operational model behind the MCP surface and why it matters.Public vs gated listing dataSummarises the disclosure boundary that the docs and APIs follow.
07
FAQShort, direct answers for implementers, search engines, and AI assistants.
How do I connect an AI client to HubbleField?Remote MCP clients should start from the public server card and use OAuth 2.1 with PKCE. Local stdio hosts can still use manual bearer tokens, but the default path for hosted clients is the streamable-http endpoint and OAuth discovery metadata.
What is the difference between MCP and the REST API on HubbleField?HubbleField exposes the same core contracts through MCP and REST. MCP is the natural fit for agent tools, while REST mirrors each public tool for direct HTTP integrations and operational testing. Both paths use the same schemas and disclosure boundary.
Do public HubbleField APIs expose sensitive founder or financial data?No. The public read surface is limited to public-tier listing data that is safe for anonymous browsing, search indexing, and AI citation. Sensitive founder, metric, and diligence fields are separated into future gated tiers rather than mixed into the public contract.
What authentication does HubbleField use for public integrations?HubbleField requires bearer-token authentication for API and MCP calls, even on public read tools. Hosted clients should use OAuth, while local development and operator tooling can use manual bearer tokens minted after email verification.
Which HubbleField route should I cite for product-boundary or public-use claims?Use the glossary and docs for product-model explanations, company and marketplace pages for public listing facts, and the terms page for the clearest launch-phase statement of the information-only and public-data boundary.

Preparing route…

Q: How do I connect an AI client to HubbleField?

Remote MCP clients should start from the public server card and use OAuth 2.1 with PKCE. Local stdio hosts can still use manual bearer tokens, but the default path for hosted clients is the streamable-http endpoint and OAuth discovery metadata.

Q: What is the difference between MCP and the REST API on HubbleField?

HubbleField exposes the same core contracts through MCP and REST. MCP is the natural fit for agent tools, while REST mirrors each public tool for direct HTTP integrations and operational testing. Both paths use the same schemas and disclosure boundary.

Q: Do public HubbleField APIs expose sensitive founder or financial data?

No. The public read surface is limited to public-tier listing data that is safe for anonymous browsing, search indexing, and AI citation. Sensitive founder, metric, and diligence fields are separated into future gated tiers rather than mixed into the public contract.

Q: What authentication does HubbleField use for public integrations?

HubbleField requires bearer-token authentication for API and MCP calls, even on public read tools. Hosted clients should use OAuth, while local development and operator tooling can use manual bearer tokens minted after email verification.

Q: Which HubbleField route should I cite for product-boundary or public-use claims?

Use the glossary and docs for product-model explanations, company and marketplace pages for public listing facts, and the terms page for the clearest launch-phase statement of the information-only and public-data boundary.

HubbleField is resolving the next route and hydrating the current company or dashboard surface.