# LangWatch (slim)

> LangWatch is an open-source platform for tracing, evaluating and testing LLM applications and agents, with prompt management, datasets and an AI gateway. It runs hosted or self-hosted, with a REST API, SDKs, a CLI and an MCP server.

- Full: https://www.anchorterminal.com/tools/langwatch.md (~8,500 tokens) · this version ~1,880 tokens · JSON https://www.anchorterminal.com/tools/langwatch.json · canonical https://www.anchorterminal.com/tools/langwatch
- Index: https://www.anchorterminal.com/llms.txt · API: https://www.anchorterminal.com/api/v1/index.json · Updated: 2026-10-09

**B · 65.5/100 · rank #291 of 842 · #7 in Agent observability & evals · not agent-ready · confidence medium**


Assessment: API keys can be limited to read or write per permission category, expire, and be revoked, and the remote MCP server uses OAuth with PKCE. The MCP server registers 101 tools with no read-only or destructive annotations, and no rate limit for the platform API was found in the reviewed documentation.

## Facts

- Kind: HTTP API · vendor: Reasoning Engine B.V. (LangWatch) · category: Agent observability & evals · legal entity: Reasoning Engine B.V. (doing business as LangWatch) · provenance 77/100
- Endpoint: `https://app.langwatch.ai` (HTTP, Streamable HTTP, SSE (legacy), stdio)
- Auth: OAuth or key · pricing: Freemium · x402: no · licence: Apache 2.0 for the platform, with an Enterprise licence for the `platform/app/ee` directory. The SDKs and the MCP server are MIT
- Probe metrics: not measured yet (probes haven't run)
- Surface graded: LangWatch Cloud at https://app.langwatch.ai, through the REST API and the MCP server. The same image runs self-hosted under Apache 2.0
- REST API: OpenAPI 3.1, 215 paths and 328 operations, 9 marked deprecated. Fields in lower_snake_case, cursor paging with `next_cursor`, errors with a stable `error.code`
- MCP server: `@langwatch/mcp-server` 2.1.0 (12 September 2026), MIT. stdio by default, or hosted at https://app.langwatch.ai/mcp (Streamable HTTP) and `/sse`. 101 tools in the source, no toolsets, no annotations
- Credentials: API keys `sk-lw-` (personal or service, All or Restricted per category, scope of projects, teams or organisation, optional expiry, revocable), write-only ingestion keys `ik-lw-`, OAuth with PKCE for remote MCP, device-code login for the CLI
- Headers: `X-Auth-Token: sk-lw-...` or `Authorization: Bearer sk-lw-...`. AI Gateway inference endpoints take a virtual key on the gateway host
- Rate limits: No platform API rate limit found in the reviewed documentation. Virtual keys on the AI Gateway take customer-set requests per minute and per day
- SDKs and CLI: Python `langwatch` 1.4.0 (6 September 2026), TypeScript `langwatch` 1.19.0 on npm (also the CLI, Node 20 or newer per the package), Go SDK 1.0.0 (9 September 2026). OpenTelemetry ingestion
- Free tier: Developer plan, no card. 50,000 events a month, 2 users, 3 scenarios, 3 simulations and 3 custom evaluations. Trace content older than 14 days is redacted
- Retention: 49 days by default per the docs. Paid plans set 35 or 63 days, Enterprise any whole number of weeks from 49 days. The pricing page says 30-day retention is included on Growth
- Audit: Enterprise plan or licence only. Logs settings, API key, member and gateway changes with user and IP address. Sign-ins and reads of product data are not logged
- Status: status.langwatch.ai on Better Stack, 9 components with 90 days of history. App 99.983 per cent, Collector 99.977 per cent and Scenarios 96.488 per cent over the period shown on 9 October 2026
- Data location: LangWatch Cloud stores data in the EU on AWS per the pricing FAQ and trust centre. Other regions on request. Self-hosted and hybrid deployments keep data on the customer's side
- Certifications: ISO 27001 and GDPR compliance stated on the trust centre and in the docs, with reports and a CREST penetration test on request. The trust centre lists SOC 2 as via AWS
- Prices: Instant Evals, judge input tokens $0.0546 per 1M tokens

- Scores: Reliability 53, Performance pending, Schema & documentation 87, Agent ergonomics 67, Security & auth 71, Payments & pricing 40, Task success pending, Maintenance & community 90, Transparency & trust 75 · negative events -2 · total over the 7 assessed categories
- Why: Reliability, Graded as a hosted service (LangWatch Cloud). · Schema & documentation, A public OpenAPI 3.1 document of 215 paths and 328 operations, served without a key, and typed zod inputs on every MCP tool (25). · Agent ergonomics, The MCP server registers 101 tools in the 2.1.0 source, with no toolsets or dynamic loading found. · Security & auth, OAuth authorisation code with PKCE and dynamic client registration on the remote MCP server, and API keys with read or write per category, a… · Payments & pricing, No x402, MPP or L402 (0). · Maintenance & community, Platform 3.20.1 on 2 October 2026, seven days before the check (30). · Transparency & trust, Apache 2.0 platform with the `platform/app/ee` directory under a separate Enterprise licence, and MIT SDKs and MCP server (27).
- Sources: 33, open questions: 8, both in the full twin
- Capabilities: obs.traces, obs.evals, obs.prompts, obs.datasets, obs.gateway
- JSON: https://www.anchorterminal.com/api/v1/tools/langwatch.json
- Verify (for the vendor): the badge `https://www.anchorterminal.com/badges/langwatch.svg` or a link to https://www.anchorterminal.com/tools/langwatch from a page on langwatch.ai or one of its subdomains, or the README of github.com/langwatch/langwatch, then `POST https://www.anchorterminal.com/api/v1/verify` `{"slug", "url"}` or `verify_listing` at /mcp; re-checked weekly, no effect on the grade. Snippets in the full twin.

## Before you call it

1. Create a Restricted key with read access to only the categories the task needs. A personal key with All permissions carries everything its owner can do
2. Set both `LANGWATCH_API_KEY` and `LANGWATCH_PROJECT_ID` for the MCP server unless the key reaches only one project
3. Call `discover_schema` before `search_traces` or `get_analytics`, and keep the default `digest` format. `json` returns the full raw trace
4. Allowlist MCP tools in the client. All 101 load by default, among them `platform_create_api_key` and the delete tools
5. Follow `next_cursor` until it is null on list endpoints. A full page does not mean more rows exist

## Connect

```bash
npm install -g langwatch
```

```bash
curl https://app.langwatch.ai/api/gateway/v1/openapi.json
```

```bash
claude mcp add langwatch --env LANGWATCH_API_KEY=sk-lw-... --env LANGWATCH_PROJECT_ID=your-project-id -- npx -y @langwatch/mcp-server
```
Full config and headless snippets are in the full page. Through letme (picks today, calling later): https://letme.dev/langwatch

## Similar tools

| Tool | Grade | Score | Shared capabilities | Slim |
| --- | --- | --- | --- | --- |
| LangSmith API + MCP | BB | 71.1 | obs.traces, obs.evals, obs.prompts, obs.datasets, obs.gateway | https://www.anchorterminal.com/tools/langsmith.min.md |
| Respan API + MCP | B | 65.5 | obs.traces, obs.evals, obs.prompts, obs.gateway, obs.datasets | https://www.anchorterminal.com/tools/respan.min.md |
| Braintrust API + MCP | C | 61.1 | obs.traces, obs.evals, obs.prompts, obs.gateway, obs.datasets | https://www.anchorterminal.com/tools/braintrust.min.md |
| Helicone AI Gateway + MCP | D | 46.9 | obs.traces, obs.gateway, obs.prompts, obs.datasets, obs.evals | https://www.anchorterminal.com/tools/helicone.min.md |
| Arize Phoenix | BB | 75.4 | obs.traces, obs.evals, obs.prompts, obs.datasets | https://www.anchorterminal.com/tools/arize-phoenix.min.md |


## Panel reviews (0, desk reviews from public material, no calls made)