The Lookback MCP server enables access to Lookback content through you favourite LLM interface for analysis. Hook it up and experience the power of agent powered analysis.
Configuration
The Lookback MCP server is available at https://mcp.lookback.io/mcp.
Here's how to configure the MCP server with various providers.
You need to be logged into your Lookback account when you configure your LLM.
After logging out of your Lookback account, your LLM will still be connected unless you disconnect it.
ChatGPT
Here's how to configure the MCP for use in ChatGPT
1. Go to ChatGPT.com
2. Click your account settings
3. Select Apps > Advanced Settings > enable developer mode
4. Click "Create App"
5. Fill the form like the screenshot below, and click "Create".
6. Accept the consent dialogue.
And you are done!
Claude Web
Here's how to configure the MCP for use on claude.ai.
Note: If you have a Claude team subscription the admin of your team might have to follow these instructions.
Click the + icon and then "Add connectors".
2. Click "Manage connectors".
3. Click "Add custom connector".
4. Enter the details as follows and click "Add"
5. Click "Connect" to connect the MCP and authenticate with Lookback.
6. Verify that the MCP works by asking Claude to list your Lookback projects.
Claude Code
Here's how to configure the MCP for use with Claude Code on the command line.
Run
claude mcp add --scope user --transport http lookback https://mcp.lookback.io/mcp.Start
claude.Enter
/mcp.Select the newly added Lookback MCP.
Select authenticate and follow the authentication steps.
Verify that the MCP works by asking Claude to list your Lookback projects.
Cursor
Here's how to configure the MCP for use with Cursor.
1. Open Cursor settings and go to Tools & MCP.
2. Click Add Custom MCP.
3. Enter the following and save.
{
"mcpServers": {
"lookback": {
"url": "https://mcp.lookback.io/mcp"
}
}
}
4. Restart Cursor
5. Go back to Tools & MCP
6. Click Connect and follow instructions
Usage
For now the MCP server can read your Lookback data, but not modify it.
It can:
Navigate within your Lookback data by listing projects, rounds, recordings, stakeholder goals, findings, tasks, and notes.
Uncover insights by searching across transcripts, findings, Eureka moments, and notes within a project.
Dive deep in analysis by accessing transcripts and transcript summaries from your sessions.
Examples
There are many ways to leverage the MCP server for post-session analysis, here are some concrete suggestions.
Thematic Analysis based on session transcripts
The MCP server facilitates access to session transcript and as such can be used to perform thematic analysis across one or more sessions.
Prompt: Perform thematic analysis on all the sessions in the "Interview" round within the "Spring 2026 Shopping" project.
What happens:
The agent fetches all the session in the "Interview" round.
The agent fetches the transcript for each session.
Based on each session transcript the agent performs thematic analysis, surfacing common themes across them.
If a less detailed approach is desired ask the agent to use the transcription summaries rather than the full transcripts.
Stakeholder Goal Coverage Audit
The MCP server enables access to Stakeholder Goals and Findings, whether created by humans or Eureka. This can be used to perform coverage audits, ensuring all Stakeholder Goals have been addressed by the research.
Prompt: Verify coverage of all Stakeholder Goals by looking through all findings in the "Spring 2026 Shopping" project and their associated goals. Highlight goals with poor coverage and findings not associated with any Stakeholder Goals.
What happens:
The agent fetches the stakeholder goals in the project.
The agent fetches the findings in the project.
The agent performs the coverage analysis and provides a summary to the user.
Surface pain points
The MCP server supports searching across all content within a project. This can be use to find and analyse specific problems or topics.
Prompt: Find pain-points around the checkout experience in the "Spring 2026 Shopping" project. Categories these by kind and provide a comprehensive overview, highlight low-hanging fruit (easy to fix, high impact).
What happens:
The agent uses the semantic search capabilities to search across transcripts, notes, and findings. It will reword the search query to find matches even if the exact terms "checkout experience" or "pain-point" aren't present.
The agent searches, potentially several times, to gather context and then carries out the review based on these results.