LangChain documents a LangSmith Model Context Protocol server that lets MCP clients read conversation threads, prompts, runs and traces, datasets, experiments, and billing usage from a LangSmith workspace. For LangSmith Cloud, docs recommend the OAuth-authenticated LangSmith Remote MCP (regional endpoints on api.smith.langchain.com and documented EU/APAC/AWS variants) with the same tool surface and no separate deployment. The standalone HTTP server at https://langsmith-mcp-server.onrender.com/mcp remains documented for API-key access via the LANGSMITH-API-KEY header, while self-hosted LangSmith users can run the open-source langsmith-mcp-server package with uvx, Docker HTTP on port 8000, or point LANGSMITH_ENDPOINT at private instances. Tools include get_thread_history, list_prompts, fetch_runs (with FQL filters and character-budget pagination), dataset/example readers, list_experiments, and get_billing_usage per the official tool table.
Use cases
- Let coding agents pull the latest failing root runs from a production project before proposing fixes
- Compare prompt templates stored in LangSmith without copy-pasting secrets into chat logs
- Page through long thread histories with documented character-based pagination limits
- Inspect experiment metrics (latency, cost, feedback) tied to a dataset before approving rollout
- Audit trace billing usage for a date range when finance asks for utilization evidence
Key features
- Cursor
- VS Code
- Claude Desktop
- Claude Code
- Codex
Frequently Asked Questions
- Should new LangSmith Cloud users use this hosted Render URL?
- Official docs mark the standalone server page as deprecated for Cloud in favor of LangSmith Remote MCP OAuth endpoints; keep Render or uvx paths for self-hosted or legacy setups.
- Which environment variables matter locally?
- LANGSMITH_API_KEY is required; LANGSMITH_WORKSPACE_ID disambiguates multi-workspace keys; LANGSMITH_ENDPOINT targets self-hosted or regional APIs.
- How does pagination work on large traces?
- get_thread_history and fetch_runs (when trace_id is set) use character-budget pages with page_number and total_pages—docs default max_chars_per_page around 25000.
Related
Related
3 Indexed items
Datadog MCP Server
Datadog documents a remote Model Context Protocol server at docs.datadoghq.com/bits_ai/mcp_server that connects AI agents in Cursor, Claude Code, Codex CLI, VS Code, Gemini CLI, and other MCP clients to observability data across APM, logs, metrics, monitors, dashboards, and security signals. Setup guides describe OAuth-based connection to Datadog's hosted MCP endpoint (distinct from the local-only Code Security MCP Server used for SAST/SCA scans). Fair-use limits listed in docs include 50 requests per 10 seconds burst and 50,000 monthly tool calls; Audit Trail records MCP actions with tool name, arguments, user identity, and client, while metrics `datadog.mcp.session.starts` and `datadog.mcp.tool.usage` tag usage by client and tool.
PostHog MCP Server
PostHog documents a free hosted Model Context Protocol endpoint at `https://mcp.posthog.com/mcp` per posthog.com/docs/model-context-protocol that lets MCP clients query analytics, manage feature flags, investigate errors, run HogQL, triage support workflows, and configure CDP destinations from natural-language prompts. The PostHog Wizard installs the server into Cursor, Claude Code, Claude Desktop, Codex, VS Code, Windsurf, and Zed via `npx @posthog/wizard@latest mcp add`. Authentication routes to the correct US or EU data region; manual setups can pass a personal API key with the MCP Server preset through `mcp-remote` and an `Authorization` Bearer header. Source lives in the PostHog monorepo at `services/mcp` (the standalone PostHog/mcp repository redirects there).
Weights & Biases MCP Server
Weights & Biases documents a hosted Model Context Protocol server at `https://mcp.withwandb.com/mcp` (recommended) and an open-source `wandb/wandb-mcp-server` package for local stdio or HTTP per docs.wandb.ai/platform/mcp-server. Authenticated clients pass a W&B API key in the `Authorization: Bearer` header. Documented tools include `query_wandb_tool` and `get_run_history_tool` for experiment metrics, `query_weave_traces_tool` and `count_weave_traces_tool` for LLM traces, `create_wandb_report_tool` for markdown reports, `search_wandb_docs_tool` for docs.wandb.ai, and `query_wandb_entity_projects` for project listings. Dedicated/on-prem deployments can set `WANDB_BASE_URL` with the local server per README.