{MCP} Pro & Team · Nugget MCP Server

Your customer research,
native to every AI agent.

Plug Claude, ChatGPT, Codex, Cursor, Windsurf, and Claude Code directly into your interview library. Specs cite real customers. Search runs over thousands of nuggets in one prompt. No copy-paste. No exports. No hallucinated users.

Start free trial — MCP on Pro 30-second setup
~/.config/claude_desktop_config.json
1{
2  "mcpServers": {
3    "nugget": {
4      "url": "https://mcp.nggt.ai",
5      "headers": {
6        "Authorization": "Bearer ngt_live_..."
7      }
8    }
9  }
10}
Connected · 38 tools available
How it works

One server. Every agent.

Nugget hosts the MCP server. Your AI clients connect over a secure tunnel using a scoped, revocable token. The server exposes your interviews, nuggets, themes, priorities, and specs as searchable, structured tools.

Claude
Claude Desktop
macOS · Windows
ChatGPT
ChatGPT
web · desktop
Cursor
Cursor
IDE
Claude Code
Claude Code
CLI
Codex
Codex
CLI · cloud
Windsurf
Windsurf
IDE
Nugget MCP
mcp.nggt.ai
38 tools online
Interviews
transcripts · audio
Nuggets
tagged + scored
Themes
clustered patterns
Specs & Priorities
structured PRDs
In practice

What your AI actually does with it.

These aren't hypotheticals. They're prompts that work the moment you connect.

USE CASE 01 / research synthesis
> What are the top three pain points enterprise customers raised in the last 30 days?
Across 14 enterprise interviews, three themes dominate:
  1. Compliance exports (mentioned 9×) — security teams need transcripts in their review systems.
  2. SSO & SCIM (7×) — onboarding friction at the IT layer.
  3. Audit trails (5×) — who changed what, when.
14 interviews 42 nuggets
USE CASE 02 / spec drafting
> Draft a PRD for the export feature using only quotes from interviews tagged "compliance".
PRD: Workspace-level Export. Targeting Q2 release. Built from 9 interview citations across 6 enterprise customers.
"It's a deal-breaker for our security review."
— VP Eng, Globex · 04/18
[Theme: Compliance] [9 quotes]
USE CASE 03 / ticket grounding
> Open a Linear issue for the highest-severity onboarding nugget. Attach the source quote.
Created ENG-2417: "First-run flow times out for SSO users." Severity: High. Linked to Nugget #1183.
↳ pushed to Linear · 2.3s
USE CASE 04 / in-IDE evidence
> While I'm coding this onboarding endpoint, surface customer quotes that mention onboarding pain.
Found 6 quotes. Top one:
"I gave up after the third screen — it just kept asking for things I already typed."
— PM, Initech · 02/04 — > cursor inline
~ 30s

Connect in three steps.

No local server, no Docker, no fiddling. Paste once, restart your client, you're in.

01
STEP 01

Generate a token

Settings → MCP. Click "New token". Choose what it can read. Tokens are read-only by default and revocable any time.

ngt_live_3f9c21a8b...
02
STEP 02

Paste the config

Drop the snippet into your client's MCP config — Claude, ChatGPT, Codex, Cursor, Windsurf, or any compatible client.

"url": "https://mcp.nggt.ai"
03
STEP 03

Restart & ask

Your client picks up the server on launch. Ask anything about your research — citations come back automatically.

38 tools online
The toolset

Thirty-eight tools. Ten domains.

The MCP server exposes Nugget as a structured API your agent can call directly. Search, read, draft, generate — every operation comes back with the source attached.

Interviews

3 tools

Recordings, transcripts, metadata.

  • interviews_list — List interviews
  • interviews_get — Get a specific interview
  • interviews_search — Search interviews

Nuggets

3 tools

Tagged signals from every conversation.

  • nuggets_list — List nuggets
  • nuggets_get — Get a specific nugget
  • nuggets_semantic_search — Semantic search

Participants

2 tools

The people behind the quotes.

  • participants_search — Search participants
  • participants_get — Get a specific participant

Priorities

4 tools

Ranked opportunities, ready to ship.

  • priorities_list — List priorities
  • priorities_get — Get a priority
  • priorities_generate — Generate priorities
  • priorities_dismiss — Dismiss a priority

Reports

5 tools

Findings packaged for stakeholders.

  • reports_list — List reports
  • reports_get — Get a report
  • reports_create — Create a report
  • reports_update — Update a report
  • reports_generate — Generate a report

Scripts

4 tools

Interview guides, kept versioned.

  • scripts_list — List scripts
  • scripts_get — Get a script
  • scripts_create — Create a script
  • scripts_update — Update a script

Specs

6 tools

PRDs grounded in real evidence.

  • specs_list — List specs
  • specs_get — Get a spec
  • specs_create — Create a spec
  • specs_update — Update a spec
  • specs_delete — Delete a spec
  • specs_generate — Generate a spec

Strategic Goals

2 tools

The north star your AI aligns to.

  • strategic_goals_list — List goals
  • strategic_goals_replace — Replace goals

Tasks

4 tools

Dev-ready work items, scoped for agents.

  • tasks_create — Create a task
  • tasks_update — Update a task
  • tasks_delete — Delete a task
  • tasks_generate — Generate tasks

Themes

5 tools

Clustered patterns across interviews.

  • themes_list — List themes
  • themes_get — Get a theme
  • themes_generate — Generate themes
  • themes_update — Update a theme
  • themes_delete — Delete a theme
Start free trial — get a token
Security & control

Your data, your leash.

Scoped tokens

Read-only by default. Limit by workspace, project, or tag.

Expiring

Set TTL per token. 24h, 30 days, or custom.

Revoke instantly

One click. Effective globally within seconds.

Audit log

Every tool call logged with token + timestamp.

Common questions

Do I need to host anything myself? +

No. The server is hosted at mcp.nggt.ai. You only paste a config snippet into your client.

Which clients are supported? +

Claude, ChatGPT, Codex, Claude Desktop, Claude Code, Cursor, Windsurf, and any client implementing the Model Context Protocol spec. If it speaks MCP, it works.

Can the agent modify my data? +

Tokens are read-only by default. You can opt-in to write scopes per-token (e.g. drafting specs back into Nugget) — never the default.

Is MCP available on the Free tier? +

No. MCP is a Pro & Team feature. Pro starts at $19/mo with a 14-day free trial.

What's the latency like? +

Tool calls return in 200–800ms for most queries. Full-corpus semantic search runs against a vector index — no cold-start, no model load on your end.

→ MCP.NGGT.AI

Stop pasting transcripts.
Start shipping evidence.

14 days free. $19/mo after. MCP included on Pro and Team.

Start free trial See pricing