TL;DR
- Live today on the ChatGPT App Directory and Claude connectors, plus Cursor, Codex, and any other MCP client.
- Ask a question in plain language. Observability’s tools return structured data or an inline chart.
- One connector, mcp.bitmovin.com, covers every Bitmovin product. Observability’s tools live under observability_*.
- Three Observability skills build on the MCP for collector integration, root cause analysis, and ad reporting.
Someone on a call asks why buffering spiked on mobile in Germany yesterday afternoon. The old answer is to open a dashboard, find the right filters, and get back to the call ten minutes later with a screenshot. That gap between the question and the answer is where most teams lose time, and it is usually a specialist who ends up closing it.
Bitmovin’s Observability solution now has an MCP Server that closes that gap directly, live on both the ChatGPT App Directory and the Claude connectors. Ask the same question in plain language from Claude, ChatGPT, or any other MCP client, and get back structured data or an inline chart without leaving the conversation, no dashboard tab required.
What it is
The Observability MCP Server lets any MCP compatible AI tool, Claude, ChatGPT, Cursor, Codex, or another MCP client, query your Observability for video playback data directly and render interactive charts inline. Ask a question in plain language and it gets translated into a precise Analytics API call behind the scenes. The result comes back as structured data that the assistant can interpret, summarize, or plot.
The recommended path is the unified mcp.bitmovin.com connector, a single MCP server that exposes tools across every Bitmovin product, Encoding, Player, Observability solution, Stream Lab, and Docs. Observability’s tools live under the observability_* namespace, and any new tool a product team ships appears automatically, no reconfiguration needed.
Authentication runs through Bitmovin SSO via OAuth. Log in once with your dashboard credentials and every product’s tools become usable, with no copying API keys between tools. An API key header is also supported for teams that prefer it.
Why it’s useful
This is conversational, self-serve access to playback QoE data, with charts appearing in the chat window instead of a separate dashboard tab. One connector covers every product, and it’s the same MCP surface that powers Bitmovin’s Observability skills and agent workflows. Advertising data, fill rate, ad abandonment, and VAST or ad server error breakdowns, plus ad alerts, are exposed through the MCP too, so the ad ops side of the house gets the same natural language access as playback QoE.
Step 1. Connect your tool
There are three ways in, pick whichever matches what you already use. All three land on the same MCP server, so the data and tools are identical regardless of path.
- ChatGPT. Go to chatgpt.com/apps, search Bitmovin, click Connect, and sign in with your Bitmovin account.
- Claude (Desktop/Web). Add mcp.bitmovin.com as a custom connector under Customize, Connectors, Add, then complete the OAuth login.
- Claude CLI and coding agents. Run claude mcp add -s user bitmovin — npx mcp-remote https://mcp.bitmovin.com/mcp, then authenticate.
API key auth is also supported for the CLI path. Append –header “x-api-key: $BITMOVIN_API_KEY” instead of authenticating with OAuth.
Step 2. Ask your first question
Once connected, just type a question the way you would ask a colleague. Here are a few to try.
- What was the 95th percentile video startup time for mobile users in Germany yesterday?
- Compare rebuffer percentage by CDN provider across the top 5 countries over the last 7 days.
- Show me a plot of error rate over the last 7 days.
- Analyze impression abc123, walk me through what happened during that session.
- Based on last week’s data, what should I improve this week?
Step 3. Go further with Observability skills
Three portable agent skills build on the Observability MCP and are worth reaching for once the connector is set up. All three are live today.
- bitmovin-observability-collector-integration. Guides setup of the Observability collector on Web, Android/Fire TV, iOS/tvOS/visionOS, and Roku, for Player and third party engines alike.
- bitmovin-observability-root-cause-analysis. Runs a structured root cause analysis on a QoE metric that moved, such as startup time, rebuffer ratio, or error rate.
- bitmovin-observability-ad-report. Produces a snapshot ad operations report, covering the impression to completion funnel, advertiser rankings, and VAST error breakdowns.
The collector is already pre-integrated for Player and takes about 3 lines of code to add to other players on web. The root cause skill also covers bitrate and video start failures, and the ad report adds pod position cohorts and ad startup latency flags.
Install with npx @bitmovin/skills, or inside Claude Code with /plugin marketplace add bitmovin/skills.
Why this matters for streaming video
Data access is moving from dashboards and query languages toward natural language and agents. That shift removes the analyst and developer bottleneck and lets PMs, ops, support, and QA self-serve answers instead of filing a ticket or waiting on a specialist. It also reflects a broader pattern across the industry, where the same interface a person uses is increasingly what an autonomous agent uses too, agents as users, as more vendors add an AI layer.
The primary audience is developers and video, QoE, and ops engineers at streaming services and broadcasters. PMs and data analysts who want answers without writing SQL or opening a dashboard are a close second, along with support and QA. Technical decision makers evaluating AI native tooling are a third group worth writing for.
Teams sit on huge volumes of playback telemetry that’s often gated behind dashboards and specialists. Conversational and agentic access shortens the loop from the dashboard turned red to knowing which CDN to fail over from, and it meets users inside the AI assistants they’re already working in. For more on what MCP is and where it fits alongside existing APIs and dashboards, see Bitmovin’s explainer on MCP for agentic AI workflows, linked below.
Availability
The Observability MCP Server is generally available today, live on both the ChatGPT App Directory and the Claude connectors. Any customer can connect via OAuth with their Bitmovin login and start asking questions right away.
Learn more
- Main documentation. https://developer.bitmovin.com/playback/docs/bitmovin-observability-mcp-server
- Observability Assistant, the in-dashboard companion. https://developer.bitmovin.com/playback/docs/the-bitmovin-observability-assistant
- Dashboard metrics reference. https://developer.bitmovin.com/playback/docs/analytics-dashboard-metrics-reference
- Skills repo. https://github.com/bitmovin/skills
- ChatGPT apps directory. https://chatgpt.com/apps
- Understanding MCP for Agentic AI Workflows, Bitmovin’s explainer on what MCP is and when to use it. https://bitmovin.com/blog/understanding-mcp-agentic-ai-data-access/
- How to Build a Streaming Platform in a Day, a hands-on walkthrough of an AI coding agent paired with Bitmovin’s MCP and CLI. https://bitmovin.com/blog/how-to-build-video-streaming-platform-ai-bitmovin/
