deeptrue MCP
OfficialClick on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@deeptrue MCP list my active projects and their current status"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
deeptrue MCP
Connect Cursor, Claude, ChatGPT, and Gemini to deeptrue surveys and research. This Model Context Protocol server reads and updates projects, briefings, surveys, fielding, panels, and reports through https://api.deeptrue.ch.

Connect in one step
Create a key under Settings → MCP API keys on app.deeptrue.ch. The secret starts with dt_ and is shown once. Node.js 20.9 or newer is required for the local command.
Paste this into the client’s MCP config (Cursor’s project or user mcp.json, or the same shape in Claude). Replace dt_replace_me. The same file is mcp.example.json.
{
"mcpServers": {
"deeptrue": {
"command": "npx",
"args": ["-y", "github:deeptrue-ag/deeptrue-mcp"],
"env": {
"DEEPTRUE_API_BASE": "https://api.deeptrue.ch",
"DEEPTRUE_API_KEY": "dt_replace_me"
}
},
"deeptrue-remote": {
"url": "https://mcp.deeptrue.ch/mcp"
}
}
}Use deeptrue when the client starts a local process. Use deeptrue-remote when it asks for a URL. You only need one of them. DEEPTRUE_API_KEY is the only value you must set. DEEPTRUE_API_BASE is optional and already points at https://api.deeptrue.ch.
Keep the key in that local file or a password manager. Do not commit it.
Related MCP server: Agent Construct
Agent skill
Cursor loads .cursor/skills/deeptrue-mcp/SKILL.md when connecting a client, calling these survey and research tools, or debugging MCP config.
Clients
Client | Connection | Credentials |
Cursor | Local command |
|
Claude | Local command, or remote | API key locally. OAuth on the remote URL, or an API key if Claude asks for one |
ChatGPT | Remote | OAuth only. Do not paste a |
Gemini | Local command if Gemini runs a process, or the remote URL | API key, or a bearer this server exchanges for a session. No separate Gemini API |
Cursor
Add the deeptrue block to Cursor’s MCP settings. Cursor starts the process and passes DEEPTRUE_API_KEY. With no arguments the command speaks MCP over stdio.
Claude
Desktop or Claude Code: the
deeptrueblock. Desktop readsclaude_desktop_config.json. Claude Code reads a project MCP file with the samemcpServersshape.Remote: add a connector with URL
https://mcp.deeptrue.ch/mcp. Sign in with deeptrue OAuth, or provide adt_…key if that Claude client asks for a header.
ChatGPT
Add a remote connector with URL https://mcp.deeptrue.ch/mcp and complete OAuth. ChatGPT remote connectors must use OAuth. Do not put a dt_… API key in the ChatGPT connector settings.
Gemini
This server has no Gemini-only connector or endpoint. If the Gemini client runs a local MCP server, use the deeptrue block and DEEPTRUE_API_KEY. If it connects to a remote MCP server, use https://mcp.deeptrue.ch/mcp.
A dt_… key is sent as an API key. Any other bearer token is sent as an OAuth access token. Both are exchanged for a session JWT before survey and project calls. You do not configure that session JWT yourself. Details are in Authentication.
Authentication
Local clients send DEEPTRUE_API_KEY. Remote clients send either of these headers:
Header | Value |
|
|
|
|
The server posts that credential to POST https://api.deeptrue.ch/v1/auth/mcp/session/ (api_key or access_token) and uses the session JWT from the response for later API calls. If the API returns 401, the server exchanges a new session once and retries the call. Revoking the API key makes the next exchange fail.
People who already have a dashboard session can also create a key:
curl -X POST https://api.deeptrue.ch/v1/auth/mcp/keys/ \
-H "Authorization: JWT <your-jwt>" \
-H "Content-Type: application/json" \
-d '{"name":"cursor"}'GET https://mcp.deeptrue.ch/healthz returns { "ok": true, "service": "deeptrue-mcp" }.
Environment variables
Variable | Required | Purpose |
| Local stdio |
|
| No | API origin. Default |
| No | Organization domain or slug. Omit it to use the organization stored on the key |
Pass organization on a tool, send X-DeepTrue-Org on HTTP, or set DEEPTRUE_ORG for a local process. The choice applies to that request. set_active_organization checks that the organization is usable; it does not stick for later remote calls.
Tools
A usual project path is create_project → submit_briefing_answers → accept_project → start_fielding → finish_fielding. Tools that change data require confirm=true and a write scope (mcp:write, or a narrower scope such as mcp:projects, mcp:surveys, mcp:fielding, or mcp:reports).
Organizations
Tool | What it does |
| Organizations available to the signed-in user |
| Address and info for one organization domain or slug |
| Check that the user can use an organization. Does not persist on stateless HTTP |
Projects
Tool | What it does |
| Projects in the active organization |
| One project by numeric id |
| Create a project. Requires a write scope |
| Patch project fields such as name or json. Requires |
| Read flattened briefing answers |
| Write briefing answers onto the project. Requires |
| Accept an offer and spawn the survey. Requires |
| Start collection for an accepted project. Requires |
| Finish a project that is collecting. Requires |
| Archive a project. Requires |
| Restore an archived project. Requires |
Surveys
Tool | What it does |
| Surveys in the active organization |
| One survey by uuid |
| Patch only |
| Statistics for a project uuid |
| Status, timing, quota, and furthest question. Personal data stays redacted unless |
Panels
These calls use deeptrue admin routes. A non-admin gets 403, and the tool reports that refusal.
Tool | What it does |
| Status snapshot for Cint, Dynata, or Bilendi |
| Cint demands or audit, Bilendi projects or audit, or Dynata projects or audit |
Reports
Tool | What it does |
| Report files for the organization. Members can call this |
| One report file by id. Members can call this |
| Layouts for audience |
| Save a report HTML snapshot. deeptrue admins only. Requires |
Architecture
Local clients start this server with npx. Remote clients open https://mcp.deeptrue.ch/mcp. Every tool calls https://api.deeptrue.ch.
flowchart LR
clients["Cursor, Claude, ChatGPT, Gemini"]
mcp["deeptrue MCP"]
api["https://api.deeptrue.ch"]
clients -->|"stdio or https://mcp.deeptrue.ch/mcp"| mcp
mcp --> apiSafety
Prefer get_project or get_survey before an edit. update_survey_structure rejects any field other than name, json, survey_template_uuid, and id. Respondent personal data needs both include_pii=true and the mcp:pii scope.
Develop
git clone https://github.com/deeptrue-ag/deeptrue-mcp.git
cd deeptrue-mcp
npm install
npm testnpm install compiles TypeScript. npx deeptrue-mcp with no arguments speaks MCP over stdio and reads DEEPTRUE_API_KEY from the environment. Node.js 20.9 or newer.
License
MIT © deeptrue AG
Available Tools
25 toolsaccept_projectAccept projectA
Accept a project offer and spawn the survey. Sequence: briefing → accept → collect. Requires confirm=true.
Inputs: project_id (required), answers, confirm, organization, +1 more.
archive_projectArchive projectBDestructive
Archive a project. Requires confirm=true.
Inputs: project_id (required), confirm, organization.
create_projectCreate projectA
Create a project in the active organization. Use when starting a new research brief. Requires mcp:write.
Inputs: name (required), json, confirm, organization, +1 more.
finish_fieldingFinish fieldingBDestructive
Finish a project that is collecting data. Requires confirm=true.
Inputs: project_id (required), confirm, organization.
get_briefingGet briefing answersBRead-only
Read flattened briefing answers JSON for a project.
Inputs: project_id (required), organization.
get_feasibilityGet panel feasibility / statusARead-only
Fetch a provider ops snapshot from the existing admin routes (cint/status, dynata/status, bilendi/projects). These routes return 403 for non-admins. There is no separate member feasibility route.
Inputs: provider.
get_organizationGet organizationBRead-only
Get organization address/info for one organization domain/slug. The choice applies only to this call.
Inputs: organization.
get_projectGet projectARead-only
Fetch a single project by numeric id. Use before lifecycle or briefing changes.
Inputs: project_id (required), organization.
get_reportGet report fileBRead-only
Fetch a report file by id.
Inputs: report_id (required), organization.
get_report_layoutsGet report layoutsCRead-only
Read project report layouts for audience admin or org.
Inputs: project_id (required), audience, organization.
get_surveyGet surveyCRead-only
Fetch a survey by uuid.
Inputs: survey_uuid (required), organization.
get_survey_statisticsGet survey statisticsCRead-only
Fetch survey statistics by project uuid.
Inputs: project_uuid (required), organization.
list_organizationsList organizationsARead-only
List organizations available to the authenticated DeepTrue user. Use when choosing an org context.
Inputs: no named parameters.
list_panel_projectsList panel projects / demandsARead-only
List Cint demands or audit, Bilendi projects or audit, or Dynata projects or audit. These routes require a deeptrue admin. A 403 is an admin refusal, not an empty list.
Inputs: limit, offset, provider, resource.
list_projectsList projectsBRead-only
List DeepTrue projects in the active organization. Use for inventory and status scans.
Inputs: limit, offset, search, organization.
list_reportsList report filesBRead-only
List report files for the active organization.
Inputs: limit, offset, organization.
list_respondentsList respondentsARead-only
List respondent progress rows for a survey (status, timing, quota, furthest question). PII redacted unless include_pii=true and token has mcp:pii scope.
Inputs: survey_uuid (required), include_pii, organization.
list_surveysList surveysCRead-only
List surveys in the active organization.
Inputs: limit, offset, organization.
restore_projectRestore projectA
Restore an archived project. Requires confirm=true.
Inputs: project_id (required), confirm, organization.
set_active_organizationCheck organizationARead-only
Validate that the user can use an organization. Does not persist across stateless HTTP requests. Pass organization on later tools or send X-DeepTrue-Org / DEEPTRUE_ORG.
Inputs: organization (required).
start_fieldingStart fieldingA
Begin data collection (collect) for an accepted project. Requires confirm=true.
Inputs: project_id (required), confirm, organization.
submit_briefing_answersSubmit briefing answersA
PATCH project.json with briefing answers. Use before accept_project. Requires confirm=true.
Inputs: answers (required), project_id (required), confirm, organization.
trigger_report_refreshSave report HTML snapshotA
POST project report HTML via projects/{id}/report/. That route is limited to deeptrue admins. Requires confirm=true. A 403 means the account is not an admin; nothing was saved.
Inputs: html (required), project_id (required), confirm, organization.
update_projectUpdate projectB
PATCH project fields (name/json/etc). Use for briefing answer edits before accept.
Inputs: patch (required), project_id (required), confirm, organization.
update_survey_structureUpdate survey structureADestructive
PATCH an existing survey with a documented survey payload only. Requires confirm=true. Do not invent schema fields.
Inputs: patch (required), survey_uuid (required), confirm, organization.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 25 tools
Most tools target clearly distinct resources and actions, with descriptions that help distinguish project, survey, briefing, fielding, and reporting operations. Minor overlap exists between update_project and submit_briefing_answers, and between get_feasibility and list_panel_projects, but these are still reasonably separable.
Tool names consistently use snake_case with a verb-first pattern such as list_projects, get_survey, create_project, and archive_project. Minor variations like set_active_organization and start_fielding still fit the same readable convention.
The server covers a broad platform spanning organizations, projects, briefings, fielding, surveys, respondents, admin feasibility, and reports, so 25 tools is plausible. However, the surface is heavy and near the threshold where tools become harder to discover and choose among.
Core project lifecycle and briefing-to-fielding workflows are well covered, and surveys and reports have read/update operations. However, notable gaps remain, such as no create/delete survey, no delete project, no respondent detail tool, and no report creation workflow.
Maintenance
Related MCP Connectors
Minds is a synthetic market research platform. This MCP server lets ChatGPT, Claude, and Cursor build Audiences from a brief, run durable Studies and Question Blocks, compare segments, and export results. OAuth 2.1 or Minds API key.
A comprehensive Model Context Protocol (MCP) server that enables AI assistants to interact with yo…
Model Context Protocol server for the Apideck Unified API. Connect any MCP-compatible agent framework to 100+ accounting systems, HRIS platforms, file storage providers, and more through one integration. More information https://www.apideck.com/mcp-server
MCP server that lets AI assistants use all OneSchema features exposed via the public API.
Related MCP Servers
- AlicenseDqualityDmaintenanceA foundation for building custom local Model Context Protocol (MCP) servers that provide tools accessible to AI assistants like Cursor or Claude Desktop.137MIT
- AlicenseNot gradedqualityDmaintenanceModel Context Protocol server that standardizes tool discovery, execution, and context management for AI applications.MIT
- AlicenseNot gradedqualityAmaintenanceEnables clients such as Claude, Cursor, VS Code, and Windsurf to manage workspaces, projects, tasks, and AI agents through a hosted OAuth endpoint or a local stdio package, and to generate MCP tools from OpenAPI specs.2,530 npmMIT
- AlicenseNot gradedqualityAmaintenanceEnables AI clients such as Claude, Cursor, VS Code, and Windsurf to list and manage workspaces, projects, tasks, AI agents, and automations through a hosted OAuth-secured endpoint or a local stdio package. Also generates MCP tools from any OpenAPI spec.2,530 npmMIT