Show HN: Crowdmind – open-source tool to test ideas against AI personas

Notifications You must be signed in to change notification settings

Fork 0

Star 3

BranchesTags

Open more actions menu

Folders and files

NameName

Last commit message

Last commit date

Latest commit

History

18 Commits

18 Commits

build

build

docs

docs

resources

resources

scripts

scripts

src

src

.gitignore

.gitignore

README.md

README.md

drizzle.config.ts

drizzle.config.ts

electron-builder.yml

electron-builder.yml

electron.vite.config.ts

electron.vite.config.ts

package-lock.json

package-lock.json

package.json

package.json

postcss.config.js

postcss.config.js

tailwind.config.ts

tailwind.config.ts

tsconfig.json

tsconfig.json

tsconfig.node.json

tsconfig.node.json

tsconfig.web.json

tsconfig.web.json

Repository files navigation

Build synthetic research panels, test ideas with AI personas, run follow-up roundtables, and export stakeholder-ready reports.

Download for Windows · Features · Templates · MCP

Why Crowdmind?

Crowdmind is a local-first desktop app for fast qualitative research before you spend money on real panels, ads, prototypes, or customer interviews.

Create a panel of synthetic personas, show them a product, message, landing page, pricing proposal, image, PDF, or multi-step funnel, and get structured reactions: scores, objections, positive signals, recurring themes, confidence indicators, and follow-up answers.

It is built for founders, product marketers, researchers, agencies, and product teams who want directional signal in minutes while keeping their research data on their own machine.

What It Does

Create realistic persona panels manually, from CSV, from marketplace templates, or with AI.

Run stimulus tests with text, multiple images, PDFs, or full funnel sequences.

Continue any test with follow-up questions, selected-persona follow-ups, or full-panel roundtables.

See the swarm directly inside the test results page, with sentiment-colored nodes and drag selection.

Chat 1:1 with any persona from their profile or directly from a test response.

Use multiple LLM providers: OpenAI, Anthropic, Gemini, OpenRouter, or the built-in offline local provider.

Export reports as JSON, Markdown, summary PDF, or a full investor/client-style PDF report.

Run locally with SQLite. Your workspaces, panels, tests, notes, and chat history stay on your machine.

Connect agentic workflows through the included CrowdMind MCP server.

Reports

Crowdmind turns raw synthetic feedback into shareable research artifacts.

PDF export supports:

Summary report: concise methodology, findings, recommendations, and limitations.

Full report: stimulus, sentiment, scorecard, recurring themes, objections, positive signals, panel voices, persona-by-persona detail, and recommendations.

Markdown and raw JSON exports are also available for Notion, docs, PRs, and custom analysis workflows.

Features

Research Panels

Workspaces and panels for organizing client, product, or market research.

AI-generated personas from a professional research brief.

Manual persona creation and “improve with AI”.

CSV import from survey/customer data.

Persona version history, so historical results stay tied to the persona version that produced them.

Persona avatars from deterministic seeds, uploads, or AI image generation.

Testing

Single stimulus tests: text, image, PDF, or multimodal.

Multi-attachment tests with multiple images and PDFs.

Funnel tests with ordered stages, drag-and-drop stage editing, and funnel templates.

Individual mode for fast parallel responses.

Focus-group mode where personas see a rotating summary of peer reactions.

Scorecards for custom criteria like clarity, trust, purchase intent, pricing fit, or usability.

Analysis

Executive summary.

Confidence and diversity badge.

Recurring themes with representative quotes.

Objections and positive signals.

Benchmarking against previous tests in the same panel.

Audience comparisons and panel timeline.

Notes with Markdown for custom analysis.

Continuation

Ask more questions after a test.

Select a subset of personas from the swarm.

Ask the full panel as a roundtable.

Use quick prompts for improving message, pricing, trust, or conversion.

Jump from any answer to the persona profile or 1:1 chat.

Providers

OpenAI

Anthropic

Google Gemini

OpenRouter

Local deterministic provider, no API key required

API keys are stored locally. When available, Crowdmind uses the OS keychain through Electron safeStorage; if OS encryption is unavailable, the Settings page makes that clear.

Getting Started

Use the App

Download the latest Windows installer:

https://github.com/Brokenwatch24/crowdmind/releases/latest

No API key is required to try Crowdmind. Use the Local provider and load the demo workspace from the welcome screen.

Run From Source

Requirements:

Node.js 20+

npm

Windows users compiling native modules need Python and Visual Studio Build Tools with Desktop C++ workload.

git clone https://github.com/Brokenwatch24/crowdmind.git cd crowdmind npm install npm run dev

Useful Commands

npm run dev # launch the Electron app in development npm run typecheck # TypeScript checks npm run smoke-test # end-to-end smoke test with the local provider npm run build # production build npm run dist:win # Windows installer in release/ npm run mcp # build and run the CrowdMind MCP server

MCP For Agentic Workflows

Crowdmind includes an MCP server for tools like Codex and other MCP-compatible agents.

npm run mcp -- --db "C:\path\to\crowdmind.sqlite"

The MCP server can list workspaces, panels, personas, generate local persona previews, save personas, run simple local tests, and fetch compact test results.

Marketplace Templates

Crowdmind ships with panel templates for:

B2B SaaS buyers

Consumer LatAm segments

Families and middle-class households

Gen Z social shoppers

Restaurant and hotel operators in Colombia

SaaS founders and operators in LatAm

You can export any panel as a .json template and share it. The app can also pull new templates directly from this repository without waiting for an app update.

To contribute one, see:

docs/MARKETPLACE_TEMPLATES.md

Architecture

Desktop shell: Electron + electron-vite

Renderer: React 18 + TypeScript + Tailwind + Radix primitives

State: Zustand

Database: SQLite with better-sqlite3 and Drizzle schema definitions

LLM layer: provider adapters with schema validation and retry/backoff

Reports: self-contained HTML rendered to PDF through Electron

MCP: stdio server for local agentic workflows

All database, filesystem, API key, and provider calls run in the Electron main process. The renderer talks through a typed window.crowdmind API exposed by the preload script.

Auto-Update

Packaged builds check GitHub Releases through electron-updater. Updates are never installed silently:

Crowdmind shows an update banner.

You click download.

After download, you choose when to restart and install.

Release assets must include the installer, blockmap, and latest.yml.

Status

Current release: v0.2.4

Verified locally with:

npm run typecheck npm run smoke-test npm run dist:win

macOS and Linux packaging are configured but not yet release-tested from their native platforms.

License

MIT

About

Desktop app for building AI persona panels and testing stimuli against them (Electron + React + TypeScript)

Topics

open-source

ai

simulation

self-hosted

swarm

ai-agents

ai-tools

mcp-server

buyer-personas

Resources

Readme

Uh oh!

There was an error while loading. Please reload this page.

Activity

Stars

3 stars

Watchers

0 watching

Forks

0 forks

Report repository

Releases 6

Crowdmind 0.2.4

Latest

Jul 12, 2026

+ 5 releases

Packages 0

Uh oh!

There was an error while loading. Please reload this page.

Contributors

Uh oh!

There was an error while loading. Please reload this page.

Languages

TypeScript 99.6%

Other 0.4%