Opik’s [MCP server](https://github.com/comet-ml/opik-mcp) connects your AI host (Claude Code, Cursor, VS Code Copilot, MCP Inspector) directly to your Opik workspace — read traces, log scores, save prompt versions, and ask Ollie investigative questions, all from the chat. ## Quick setup with the Opik CLI The fastest way to connect the MCP server is the Opik CLI. It detects your AI host (Claude Code, Cursor, VS Code Copilot), picks the right server for your Opik deployment, and configures it for you. Prefer not to use the CLI? You can wire up any host by hand — skip to [Manual setup](/content/docs/opik/mcp-server#manual-setup/index.html). ### Install the Opik CLI The CLI ships with the `opik` Python package. The `opik mcp` commands require **version 2.1.3 or later**: ``` pip install --upgrade "opik>=2.1.3" ``` ### Configure the MCP server ``` opik mcp configure ``` This reuses your existing Opik configuration (`~/.opik.config`). If you haven’t configured Opik yet, the wizard offers to do it for you first. `opik configure` also offers to run this for you at the end of its setup: > Set up the Opik MCP server for an AI assistant (Claude Code, Cursor, VS Code)? (y/N) ### Restart your AI host and verify Restart your AI host, then ask **“list my Opik projects”** in the chat to confirm it works. If your host isn’t detected, use [Manual setup](/content/docs/opik/mcp-server#manual-setup/index.html) below. ## Check your setup Each AI host keeps its own copy of the MCP configuration, which isn’t updated automatically when your Opik configuration changes. To see what every detected host points at — and whether it still matches your current Opik configuration — run: ``` opik mcp status ``` It prints your active Opik configuration, then each assistant that has the Opik MCP server configured: the config file it lives in, the server it reports to (hosted or local), its workspace, and whether it has drifted from your Opik configuration. ``` Your Opik configuration File ~/.opik.config Environment https://www.comet.com/opik/api Workspace my-workspace Opik MCP server — configured for 1 AI assistant: Claude Code Config ~/.claude.json Connection Hosted (HTTP + OAuth) Reports to https://www.comet.com/opik/api/v1/mcp Status ✓ in sync with your Opik configuration ``` A host that has drifted is flagged `✗ OUT OF SYNC` — re-run `opik mcp configure` to fix it. A host keeps its MCP connection for the lifetime of its process. After changing your Opik configuration or re-running `opik mcp configure`, **restart your AI host** so it reconnects with the updated settings. To view just your active Opik configuration (file path, environment, workspace): ``` opik configure status ``` ## Opik Cloud and self-hosted deployments `opik mcp configure` works the same whether you’re on Opik Cloud, self-hosted, or a local install — it sets up the right server for your deployment automatically. ### Opik Cloud (hosted server) On [Opik Cloud](/content/opik/index.html), the CLI registers the **hosted MCP server** over HTTP. Your AI host signs in with a browser-based OAuth flow on first connect, so: - **No API key is stored** in the host’s config — you authenticate through OAuth in the browser. - **`uv` is not required** — there is no local process to run. - Your workspace is selected during the OAuth sign-in, so a hosted server shows no workspace in `opik mcp status`. ### Self-hosted and local (local server) If no hosted server is available for your environment, the CLI sets up the **local server**, which runs on demand via `uvx opik-mcp`. This requires [`uv`](https://docs.astral.sh/uv/) — install it with `brew install uv` (macOS) or `curl -LsSf https://astral.sh/uv/install.sh | sh`. ## Manual setup Prefer to wire it up yourself, or your host wasn’t detected? Configure any host by hand below. There are two servers you can add by hand. [`opik mcp configure`](/content/docs/opik/mcp-server#quick-setup-with-the-opik-cli/index.html) picks the right one for you, but you can also add either directly in your AI host’s MCP settings: - **Hosted server** (HTTP + OAuth) — available on Opik Cloud and any deployment that provides it. No API key is stored; your host signs in through the browser. - **Local server** (`uvx opik-mcp`, stdio) — runs on your machine with your credentials in the host’s `env` block. ### Hosted server (Opik Cloud) The hosted server connects over HTTP and signs in with a browser-based OAuth flow on first connect — no API key is stored in the host config. Point your host at your deployment’s MCP endpoint, which is your Opik API base plus `/v1/mcp`. On Opik Cloud that is `https://www.comet.com/opik/api/v1/mcp`. Add the server with one command: ``` claude mcp add --transport http opik-mcp https://www.comet.com/opik/api/v1/mcp ``` Or edit `~/.claude.json` directly: ``` { "mcpServers": { "opik-mcp": { "type": "http", "url": "https://www.comet.com/opik/api/v1/mcp" } } } ``` Restart Claude Code and complete the browser sign-in when prompted, then ask in the chat: **“list my Opik projects”**. ### Local server (uvx) The local server runs on demand via `uvx opik-mcp` (requires [`uv`](https://docs.astral.sh/uv/)), with your credentials passed through the host’s `env` block. `opik-mcp` is now a Python package. If you previously ran the npx-based JavaScript server, use the `uvx opik-mcp` commands below in place of `npx -y opik-mcp`. `OPIK_WORKSPACE` is **optional** — you can omit the `OPIK_WORKSPACE` line/key entirely and the server uses the `default` workspace (correct for local/OSS installs). The snippets below include it for completeness; set it only if you connect to a named cloud workspace. Add the server with one command: ``` claude mcp add --transport stdio opik-mcp \ --env OPIK_API_KEY= \ --env OPIK_WORKSPACE= \ -- uvx opik-mcp ``` Or edit `~/.claude.json` directly: ``` { "mcpServers": { "opik-mcp": { "type": "stdio", "command": "uvx", "args": ["opik-mcp"], "env": { "OPIK_API_KEY": "", "OPIK_WORKSPACE": "" } } } } ``` Restart Claude Code, verify with `/mcp` (`opik-mcp` should appear as connected), and then ask in the chat: **“list my Opik projects”**. ## Using the MCP server ### The tools at a glance | Tool | Purpose | | --- | --- | | `read` | Universal read by id / name / `opik://` URI. | | `list` | Universal list with optional name filter and pagination. | | `ask_ollie` | Investigate or synthesize via the Opik in-product assistant. | | `write` | Universal write — log traces/spans, score, comment, save prompts, manage test suites and experiments. | | `schema` | Introspect write-operation schemas (used by the LLM to construct valid payloads). | | `run_experiment` | Run an evaluation experiment end-to-end via Ollie. | ### Browsing your workspace > list my Opik projects > what was the most recent trace logged to the “demo” project? > show me trace `` ### Scoring, commenting, saving prompts > score trace `` 0.9 on helpfulness with reason “great recovery” > comment “retry with temperature=0” on span `` > save the following text as a new version of the “rerank-system” prompt: … For the full set of write operations and their payload shapes, ask the host **“show me the schema for trace.create”** (calls the `schema` tool) or see the [README](https://github.com/comet-ml/opik-mcp#tools). ### Asking Ollie For investigative or cross-entity questions: > why are spans in the “demo” project slower this week than last? > compare experiments “rerank-v2” and “rerank-v3” on factuality `ask_ollie` returns a `thread_id` you can pass back on follow-ups to preserve context. For more about Ollie itself, see [Ollie](/content/docs/opik/ollie/index.html). See [Ollie & auto-approve](/content/docs/opik/mcp-server#ollie--auto-approve/index.html) below before running write-style prompts in shared workspaces. ## Ollie & auto-approve By default, writes that Ollie performs mid-stream (scores, comments, prompt versions, test-suite items) execute without a per-action confirmation step. Each auto-approved write is logged as a JSON audit row on the `opik_mcp.audit` Python logger. To require manual confirmation instead, set `OPIK_MCP_AUTO_APPROVE=disabled` in the server’s `env` block. Ollie’s confirmation requests then surface as typed errors that you can re-issue manually. ## Known host limits - **Cursor enforces a 60-second hard tool-call timeout** that does not reset on progress notifications. Long `ask_ollie` turns will fail on Cursor. For long-running investigations, use Claude Code or VS Code Copilot. ## Example conversation A typical investigative loop using Claude Code: > **You:** Why did the experiment “gpt-4o-rerank-v3” regress on factuality? > > **Claude:** _(calls `ask_ollie`)_ Three traces failed because the reranker dropped the system message. The remaining 12 traces scored above 0.8… > > **You:** Score the bottom 3 traces 0.2 with reason “dropped system message”. > > **Claude:** _(calls `write` with `score.create` ×3)_ Done — three scores recorded on traces ``, ``, ``.