generated: '2026-08-29' method: searched source: https://docs.spyderbat.com/installation/mcp.md name: Spyderbat MCP Server description: >- A first-party, provider-hosted Model Context Protocol server that exposes Spyderbat's search, investigation and management surface to MCP-compatible AI clients. Documented and supported by Spyderbat with one-click installers for Cursor, VS Code and LM Studio and a documented `claude mcp add` command. Transport is streamable-HTTP; authentication is an API key presented as `Authorization: Bearer ` — the same key used for the REST API, inheriting the creating user's RBAC role. The docs state plainly that OAuth is not currently supported. status: published deployment: mode: remote endpoint: https://api.spyderbat.com/mcp/v1/mcp auth: api-key verified: probed transport: streamable-http protocol_version_header: MCP-Protocol-Version regions: - region: US endpoint: https://api.spyderbat.com/mcp/v1/mcp - region: Mumbai endpoint: https://api.mum.prod.spyderbat.com/mcp/v1/mcp install_command: >- claude mcp add --transport http -s user spyderbat https://api.spyderbat.com/mcp/v1/mcp --header "Authorization: Bearer " x-evidence: fetched: '2026-08-29' tools_list_probe: url: https://api.spyderbat.com/mcp/v1/mcp method: 'POST {"jsonrpc":"2.0","id":1,"method":"tools/list"}' http_status: 403 result: >- Anonymous tools/list is rejected with HTTP 403 and no WWW-Authenticate challenge — the server exists and is auth-gated behind an API key. The response advertised MCP-Protocol-Version and Mcp-Session-Id in access-control-allow-headers, confirming a live MCP endpoint rather than a generic 403 page. oauth_discovery: - url: https://api.spyderbat.com/.well-known/oauth-protected-resource http_status: 404 - url: https://api.spyderbat.com/.well-known/oauth-authorization-server http_status: 404 note: >- No OAuth discovery documents are served, consistent with the docs statement that the MCP server authenticates by API key only. Tool inputSchemas therefore require authenticated introspection; the tools below are the ones Spyderbat names in its own documentation, with names and purposes verbatim from the docs — NOT a live tools/list dump, and NOT derived from the OpenAPI. tools: - name: onboard description: Teach the assistant what Spyderbat tools exist and how to use them. category: orientation source: docs - name: prime description: Load Spyderbat investigation workflows and schemas into the assistant's context. category: orientation source: docs - name: search_docs description: Search the Spyderbat documentation for query syntax, schemas and concepts. category: orientation source: docs - name: list_organizations description: List the organizations the calling API key has access to, with names and UIDs. category: organization source: docs - name: get_current_time description: Return the current time, used to bound and validate search time ranges. category: utility source: docs - name: get_trace_contents description: >- Pull the processes, connections and red flags linked to a Spydertrace, walking the process tree via ppuid to show parent-child relationships. category: investigation source: docs - name: get_objects description: >- Resolve object UIDs (for example a pod_uid) into full objects — pod name, namespace and node. category: investigation source: docs - name: suppress_trace description: >- Generate a suppression policy for a Spydertrace, with preview=true to render the policy and its scope before it is applied. category: response source: docs write: true reversible_via: Suppression policies are managed objects and can be deleted or edited after creation. search_schemas: note: >- The docs describe search as running against named schemas rather than as one tool per schema. Schemas named explicitly in the MCP documentation. schemas: - model_spydertrace - model_connection - event_redflag clients_documented: - Claude Code - Cursor - Windsurf - VS Code (Cline / Continue) - LM Studio - MCP Inspector