{"site":{"name":"Koji","description":"AI-native customer research platform that helps teams conduct, analyze, and synthesize customer interviews at scale.","url":"https://www.koji.so","contentTypes":["blog","documentation"],"lastUpdated":"2026-05-18T12:58:35.225Z"},"content":[{"type":"documentation","id":"19fbd450-dc0f-495a-8fb6-667b074fc6ac","slug":"mcp-overview","title":"Koji MCP Integration Overview","url":"https://www.koji.so/docs/mcp-overview","summary":"Koji MCP integration connects the AI-powered research platform to Claude and other AI assistants via the Model Context Protocol. It provides 15 tools across 5 categories (read, create, analyze, customize, distribute) enabling users to manage their entire research workflow conversationally — including creating studies with structured question types (scales, rankings, multiple choice), analyzing per-question aggregations, and generating shareable reports. Uses OAuth 2.1 with PKCE authentication and works with Claude Desktop, Cursor, and Windsurf.","content":"## What is the Koji MCP Integration?\n\nThe Koji MCP integration lets you connect your AI-powered research platform directly to Claude, Cursor, Windsurf, and any AI assistant that supports the [Model Context Protocol (MCP)](https://modelcontextprotocol.io). Instead of switching between tabs, you manage your entire qualitative research workflow through natural conversation.\n\nThink of it as giving Claude direct access to your Koji account. You can say things like:\n\n- \"Create a new customer discovery study about why users churn after onboarding\"\n- \"Add an NPS question on a 0-10 scale with anchor probing and 3 open-ended follow-ups\"\n- \"Show me the average satisfaction score and top themes from my pricing study\"\n- \"Generate a report from my latest interviews and publish it for my team\"\n\nAll without leaving your AI assistant.\n\n---\n\n## Why Use MCP Instead of the Dashboard?\n\nThe Koji dashboard gives you full visual control. MCP gives you **speed and automation**. Here is when each makes sense:\n\n| Scenario | Dashboard | MCP via Claude |\n|----------|-----------|----------------|\n| Browse studies visually | Best | Good |\n| Create a study with structured questions | Good | Best — Claude configures question types, scales, and probing |\n| Cross-analyze multiple interviews | Manual | Best — Claude synthesizes patterns |\n| Generate and iterate on reports | Good | Best — conversational refinement |\n| View per-question aggregations (NPS averages, choice distributions) | Good | Best — instant quantitative summaries |\n| Bulk import respondents | Good | Best — natural language input |\n| Customize branding in detail | Best | Good |\n| Quick status check | Requires login | Best — one question |\n\nThe real power is **combining Claude's reasoning with your research data**. Claude does not just read your data — it can create studies with structured question types (scales, multiple choice, rankings), help you refine interview questions and probing strategies, cross-reference findings across studies, and generate shareable reports.\n\n---\n\n## What Can You Do?\n\n### 15 Tools Across 5 Categories\n\nThe Koji MCP server exposes 15 tools organized into logical groups:\n\n#### Read and Explore (5 tools)\n| Tool | What It Does |\n|------|-------------|\n| `koji_list_studies` | List all your studies with status, interview counts, and completion metrics |\n| `koji_get_study` | Get full study details including research brief, structured questions, methodology, and interview data |\n| `koji_get_interviews` | List interviews with AI-generated summaries, themes, and sentiment |\n| `koji_get_transcript` | Read full interview transcripts with per-question structured answers and analysis |\n| `koji_get_account` | Check your plan, credit balance, and available features |\n\n#### Create and Publish (3 tools)\n| Tool | What It Does |\n|------|-------------|\n| `koji_create_study` | Create a study with structured questions (open-ended, scale, choice, ranking, yes/no) and methodology framework |\n| `koji_update_brief` | Modify brief sections, replace questions, change methodology, adjust probing settings |\n| `koji_publish_study` | Make a draft study live — sets up AI interviewer, generates interview link |\n\n#### Analyze and Report (4 tools)\n| Tool | What It Does |\n|------|-------------|\n| `koji_get_study_data` | Get aggregated data with per-question statistics (scale averages, choice frequencies, ranking positions) |\n| `koji_generate_report` | Generate a comprehensive AI research report from interview data |\n| `koji_get_report` | Retrieve report sections — executive summary, themes, charts, recommendations |\n| `koji_publish_report` | Publish report with a shareable public link for stakeholders |\n\n#### Customize (1 tool)\n| Tool | What It Does |\n|------|-------------|\n| `koji_configure_study` | Configure branding, lead forms, interaction modes, URL slug, and Open Graph settings — all in one tool |\n\n#### Distribute (2 tools)\n| Tool | What It Does |\n|------|-------------|\n| `koji_export_data` | Export study data as structured JSON for external analysis |\n| `koji_import_respondents` | Import contacts with personalized interview URLs for outreach |\n\n---\n\n## Structured Questions via MCP\n\nOne of the most powerful MCP capabilities is creating studies with structured question types. Instead of plain open-ended questions, you can configure:\n\n- **Scale questions** — NPS (0-10), satisfaction (1-5), Likert scales (1-7) with labeled endpoints and anchor probing\n- **Single and multiple choice** — Segment respondents with predefined options and optional \"Other\" free-text\n- **Ranking questions** — Prioritize features, pain points, or alternatives by asking respondents to rank options\n- **Yes/No questions** — Quick screening or validation checks\n- **Open-ended questions** — Deep qualitative exploration with configurable probing depth (0-3 follow-ups)\n\nThe AI interviewer presents these as interactive widgets during the conversation and captures structured data alongside qualitative context. After interviews complete, `koji_get_study_data` returns per-question aggregations (averages, distributions, frequencies) and `koji_get_transcript` includes typed structured answers for each question.\n\n---\n\n## Supported AI Assistants\n\nThe Koji MCP server works with any client that supports the Model Context Protocol:\n\n- **Claude Desktop** — Anthropic's desktop app (recommended)\n- **Claude.ai** — Browser-based Claude with MCP support\n- **Cursor** — AI-powered code editor\n- **Windsurf** — AI development environment\n- **Any MCP-compatible client** — The protocol is an open standard\n\n---\n\n## Plan and Credit Information\n\nMCP access is available on all plans. Credits are the only gate for usage:\n\n| Action | Credit Cost |\n|--------|------------|\n| Text interview | 1 credit |\n| Voice interview | 3 credits |\n| Report generation | 5 credits |\n\nAll plans include MCP access. Study creation is unlimited on all plans. Voice interviews require the Interviews plan or higher with ElevenLabs integration enabled. See the [plan comparison guide](/docs/plan-comparison-guide) for full details.\n\n---\n\n## Security and Privacy\n\n- **OAuth 2.1 with PKCE** — Industry-standard secure authentication. No API keys stored in config files.\n- **Scoped access** — MCP only accesses your studies, interviews, and account data.\n- **No training data** — Your research data is never used to train AI models.\n- **Token rotation** — Access tokens expire every hour with automatic refresh.\n- **Revocable** — Disconnect any time from your Koji account settings.\n\nLearn more in the [Authentication and Security guide](/docs/mcp-authentication-security).\n\n---\n\n## Next Steps\n\n1. **[Set up the connection](/docs/mcp-setup-claude)** — Connect Koji to Claude in under 2 minutes\n2. **[Explore the tool reference](/docs/mcp-tool-reference)** — Detailed docs for all 15 tools\n3. **[Follow a workflow guide](/docs/mcp-workflow-product-managers)** — End-to-end examples for your role\n4. **[Learn about structured questions](/docs/structured-questions-guide)** — Configure question types, scales, and probing\n\n## Further reading on the blog\n\n- [Best AI Market Research Tools in 2026: The Complete Buyer's Guide](/blog/ai-market-research-tools-2026) — AI has fundamentally changed market research. This guide compares the leading AI market research platforms—from AI-native interview tools li\n- [AI-Moderated Interview Platforms Compared: Which One Actually Works? (2026)](/blog/ai-moderated-interview-platforms-2026) — Not all AI interview platforms deliver real qualitative depth. This guide compares the top AI-moderated interview platforms in 2026 — Koji, \n- [Koji vs Remesh: AI Interviews vs AI Focus Groups — Which Research Method Wins? (2026)](/blog/koji-vs-remesh-2026) — Remesh runs AI-powered live focus groups with up to 1,000 participants. Koji runs async AI interviews with zero groupthink. Here is the full\n\n<!-- further-reading:blog -->\n","category":"Claude & MCP Integration","lastModified":"2026-05-13T00:25:38.788654+00:00","metaTitle":"Koji MCP Integration — Connect AI Research to Claude | Koji Documentation","metaDescription":"Connect Koji to Claude and AI assistants via MCP. Create studies, analyze interviews, generate reports, and automate qualitative research workflows conversationally. 17 tools, OAuth 2.1 security.","keywords":["Koji MCP","MCP integration","Claude AI research","AI interview automation","Model Context Protocol","qualitative research AI","automate user interviews","Claude MCP server","AI research assistant","product manager AI tools"],"aiSummary":"Koji MCP integration connects the AI-powered research platform to Claude and other AI assistants via the Model Context Protocol. It provides 15 tools across 5 categories (read, create, analyze, customize, distribute) enabling users to manage their entire research workflow conversationally — including creating studies with structured question types (scales, rankings, multiple choice), analyzing per-question aggregations, and generating shareable reports. Uses OAuth 2.1 with PKCE authentication and works with Claude Desktop, Cursor, and Windsurf.","aiPrerequisites":["A Koji account (free tier works for basic features)","An MCP-compatible AI assistant (Claude Desktop recommended)"],"aiLearningOutcomes":["Understand what the Koji MCP integration enables","Know which tools are available and what they do","Understand plan requirements for different features","Choose the right setup path for your AI assistant"],"aiDifficulty":"beginner","aiEstimatedTime":"8 minutes"}],"pagination":{"total":1,"returned":1,"offset":0}}