generated: '2026-07-19' method: searched status: published source: https://github.com/lightrun-platform/lightrun-ai/blob/main/server.json docs: https://docs.lightrun.com/mcp/mcp-overview/ quickstart: https://docs.lightrun.com/mcp/mcp-quickstart/ tools_reference: https://docs.lightrun.com/mcp/lightrun-tools/ server: name: com.lightrun/mcp-server title: Lightrun MCP Server description: Connect AI assistants to live runtime context for production debugging. version: 1.0.0 transport: streamable-http url: https://app.lightrun.com/mcp repository: https://github.com/lightrun-platform/lightrun-ai subfolder: lightrun-mcp-server manifest: mcp/lightrun-server.json introduced: v1.75 (January 2026) embedded_in: Lightrun Server authentication: type: oauth2 note: The MCP endpoint is served from the same host as the Lightrun app and authorizes through the Lightrun Keycloak realm. The realm advertises a dedicated `mcp` scope in its OpenID discovery document, alongside authorization_code, client_credentials, device_code and token-exchange grants with PKCE (S256). authorization_endpoint: https://app.lightrun.com/auth/realms/lightrun/protocol/openid-connect/auth token_endpoint: https://app.lightrun.com/auth/realms/lightrun/protocol/openid-connect/token scope: mcp discovery: well-known/lightrun-openid-configuration.json access_control: organization_level: true note: Organization-level MCP access control shipped in v1.86 (3 June 2026). clients: - name: Cursor config: https://github.com/lightrun-platform/lightrun-ai/blob/main/.cursor/mcp.json - name: Claude Code config: https://github.com/lightrun-platform/lightrun-ai/blob/main/lightrun-mcp-server/examples/claude-code.md - name: GitHub Copilot (VS Code) config: https://github.com/lightrun-platform/lightrun-ai/blob/main/lightrun-mcp-server/examples/copilot-vscode-mcp.json - name: OpenAI Agents config: https://github.com/lightrun-platform/lightrun-ai/blob/main/lightrun-mcp-server/examples/openai-agents.md tools: - name: get_runtime_sources description: Discover available applications, services and runtime environments available for debugging. - name: get_runtime_expression_values description: Inspect variable values and expressions at a code location (synchronous; v1.75-1.84). - name: get_runtime_callstack description: Capture call stacks at a specific code location (synchronous; v1.75-1.84). - name: snapshot_create description: Initialize an asynchronous runtime snapshot action (v1.85+). - name: snapshot_status description: Poll the status of a snapshot action. - name: snapshot_cancel description: Terminate a running snapshot action. - name: snapshot_get_values description: Retrieve captured expression values from a snapshot. - name: snapshot_get_call_stack description: Retrieve captured call stacks from a snapshot. - name: get_runtime_execution_duration description: Measure how long a code section takes to execute (synchronous; v1.78-1.86). - name: get_runtime_execution_duration_samples description: Retrieve detailed execution-duration measurement samples. - name: get_runtime_execution_count description: Count how many times a code path executes (synchronous; v1.80-1.88). - name: execution_count_create description: Initialize an asynchronous execution-count action (v1.89+). - name: execution_count_status description: Poll the status of an execution-count action. - name: execution_count_samples description: Retrieve execution-count samples. - name: execution_count_cancel description: Stop execution-count collection. - name: get_runtime_numeric_metric description: Capture integer expression values as a metric (synchronous; v1.79-1.88). - name: get_runtime_numeric_metric_samples description: Retrieve numeric-metric sample details. - name: custom_metric_create description: Initialize an asynchronous custom-metric action (v1.89+). - name: custom_metric_status description: Poll the status of a custom-metric action. - name: custom_metric_samples description: Retrieve custom-metric samples. - name: custom_metric_cancel description: Stop custom-metric collection. - name: get_actions description: List the Lightrun actions currently in place, for AI assistants (added v1.88, 30 June 2026). notes: Lightrun is shifting its MCP surface from synchronous get_runtime_* tools to create/status/samples/cancel tool quartets so long-running captures do not block the assistant. The synchronous variants remain documented with the version ranges above. deployment: mode: remote endpoint: https://app.lightrun.com/mcp verified: probed probe: gated checked: '2026-09-11' source: wall re-adjudication 2026-09-11 (RFC 9728 / JSON-RPC envelope) probe_prior: wall probe_why: RFC 9728 challenge on the MCP path only