Direct answer

To connect an MCP client to SPYN3, create a permission-scoped API key in the beta dashboard and configure the client with the hosted /agent/mcp Streamable HTTP endpoint plus a Bearer authorization header. The workspace is resolved from the credential, so the client does not send or choose a tenant identifier.

What you need

  • A SPYN3 private-beta workspace with at least one indexed source.
  • An API key containing search:knowledge for search and answers, plus read:documents if the client must list sources or retrieve document detail.
  • An MCP client that supports remote HTTP servers and custom authorization headers.
Connection plus behavior

This page configures the protocol connection. The downloadable agent-agnostic SPYN3 skill adds a portable, evidence-first operating method after the connection exists.

Keep the real values private

The beta dashboard supplies the hosted API origin and shows an API-key secret once. The placeholders below are intentional. Do not paste a real secret into source control, public issue trackers or shared screenshots.

Add the hosted server

Use the configuration shape supported by your MCP host. In clients that accept a server map, the private-beta setup is:

MCP client configurationJSON
{
  "mcpServers": {
    "spyn3": {
      "type": "http",
      "url": "${SPYN3_API_URL}/agent/mcp",
      "headers": {
        "Authorization": "Bearer ${SPYN3_API_KEY}"
      }
    }
  }
}

After the client initializes the connection, call tools/list to discover only the tools permitted by that credential. SPYN3 currently supports the MCP lifecycle methods initialize, tools/list and tools/call on the beta endpoint.

The four current MCP tools

READsearch_workspaceRun semantic and keyword search and return ranked source snippets.
READquery_workspaceAsk a natural-language question and receive an answer with supporting source records.
READlist_sourcesInspect the data sources currently connected to the workspace.
READget_documentRetrieve bounded chunks from a selected document by stable identifier.

The small surface is deliberate. Search first, then load detail only when the agent needs it. This reduces tool-selection ambiguity and protects the client's context window. The architecture is explained in why SPYN3 uses one MCP endpoint per workspace.

What the server returns

MCP tool calls return text content shaped for agent consumption. Search records include a document identifier, title, connector identifier, relevance score and bounded snippet. Answer calls return a natural-language answer followed by its source records. Provenance cannot be disabled.

Defaults can be tightened through response policy. Applications that need ordinary JSON can use the REST quickstart and request response_format: "json".

Private-beta boundary

The public documentation describes the interface implemented in the current SPYN3 beta. Access, hosted origins and credentials are issued through onboarding. Protocol-version support, OAuth availability and client-specific configuration will be published with each beta release; do not infer general availability from this page.

One endpoint should outlive one client

Connect your first real workspace

Bring one source and one agent job. SPYN3 gives the client a governed MCP surface while you keep control of prompts, approvals and application behavior.

Request private beta access
About SPYN3

SPYN3 is AI-ready data infrastructure for technical builders. It connects business sources into a workspace-isolated knowledge layer and exposes that layer through REST and a hosted MCP server, with permission-scoped access and cited retrieval. SPYN3 is currently in private beta.

Sources and implementation basis

  1. Model Context Protocol specification — protocol architecture and current standard. Verified 13 Sep 2026
  2. SPYN3 endpoint, tool and response behavior was verified against the current private-beta gateway implementation on 13 Sep 2026.