CivAgent researches a company, verifies live evidence, designs an agent operating model, runs sandbox analysis, syncs proof to Supabase, and exports board-ready artifacts.
Download · Quick Start · Workflow · Docs
Install CivAgent Desktop from the public v1.0.3 release.
| Platform | Download |
|---|---|
| macOS | CivAgent-Desktop-1.0.3-mac-arm64.dmg |
| macOS ZIP | CivAgent-Desktop-1.0.3-mac-arm64.zip |
| Windows | CivAgent-Desktop-1.0.3-win-x64.exe |
| Linux | CivAgent-Desktop-1.0.3-linux-x86_64.AppImage |
macOS builds are unsigned until Apple Developer signing credentials are added, so macOS may show a first-open security warning.
If macOS says "CivAgent Desktop is damaged and can't be opened", remove the download quarantine flag:
xattr -cr "/Applications/CivAgent Desktop.app"
open "/Applications/CivAgent Desktop.app"If you have not moved the app into Applications yet:
cp -R "/Volumes/CivAgent Desktop/CivAgent Desktop.app" "/Applications/"
xattr -cr "/Applications/CivAgent Desktop.app"
open "/Applications/CivAgent Desktop.app"CivAgent turns an organization or market target into a deployable AI-agent operating plan.
| Output | Purpose |
|---|---|
| Company intelligence brief | Live evidence, market context, and source URLs |
| Agent team map | Role-by-role agent operating model |
| Workflow blueprint | Automations, tool access, and approval gates |
| Risk register | Governance, compliance, and failure-mode planning |
| Venture memo | Pricing, moat, GTM, and scale strategy |
| Export bundle | JSON plus markdown report for sharing |
| Surface | Role |
|---|---|
| Website | Static product site at / with positioning, architecture, security notes, and download links |
| Desktop app | Electron software with a fixed command workspace, sidebar navigation, run tools, evidence panels, exports, and audit logs |
| Backend sidecar | Python service for readiness, provider calls, local evidence, exports, and Supabase sync |
The website is the front door. The desktop app is the product.
CivAgent is intentionally strict: a run is not completed unless every required provider is configured and succeeds.
| Provider | Job |
|---|---|
| Gemini | Reasoning and artifact generation |
| Tavily | Live company and market research |
| Firecrawl | Required website extraction |
| Composio | SaaS/action toolkit graph |
| E2B | Sandboxed ROI, workflow, and risk analysis |
| Supabase | Cloud evidence sync for runs, artifacts, sources, tool calls, approvals, and audit events |
Create .env from .env.example, fill every required value, then restart CivAgent:
AI_PROVIDER=gemini
GEMINI_API_KEY=...
GEMINI_MODEL=gemini-3-flash-preview
TAVILY_API_KEY=...
FIRECRAWL_API_KEY=...
COMPOSIO_API_KEY=...
E2B_API_KEY=...
SUPABASE_URL=...
SUPABASE_PUBLISHABLE_KEY=...
SUPABASE_SECRET_KEY=...For development, keep .env in the repository root. For the packaged desktop app, use the app data .env path shown inside the workspace readiness panel, for example ~/Library/Application Support/civagent/.env on macOS.
Before the first real run, open Supabase SQL Editor and run supabase/schema.sql. CivAgent will not mark a run complete until those tables exist and Supabase sync succeeds.
cd "/Users/aman_shh/Documents/New project 2"
npm install
npm run setup:python
npm run brand:icons
cp .env.example .env
# edit .env with Gemini, Tavily, Supabase, Firecrawl, Composio, and E2B values
npm run check
npm run desktop:devnpm run desktop:dev reads the repository .env file. The downloaded desktop installer reads the app data .env path shown inside CivAgent.
Server-only development:
npm startOpen:
http://localhost:8080
http://localhost:8080/app
Build desktop packages:
npm run backend:build
npm run desktop:distElectron Desktop
-> starts Python backend on 127.0.0.1
-> waits for /api/health
-> opens /app command workspace
-> reads required provider values from .env or backend environment
-> runs Gemini + Tavily + Firecrawl + Composio + E2B
-> syncs evidence to Supabase
-> stores local SQLite audit backup
-> exports markdown and JSON artifacts
Core files:
| File | Role |
|---|---|
index.html |
Public product website |
app.html / script.js |
Desktop workspace experience |
server.py |
Backend API, provider orchestration, sync, exports, audit logs |
desktop/main.js |
Electron shell and backend lifecycle |
The README diagram is GitHub-safe Mermaid. For a zoomable version, open docs/workflow-diagram.html.
flowchart LR
classDef desktop fill:#10201d,stroke:#72e0c4,color:#f7f3e8,stroke-width:1px
classDef gate fill:#2a2112,stroke:#ffb454,color:#f7f3e8,stroke-width:1px
classDef tool fill:#111827,stroke:#9fd7ff,color:#f7f3e8,stroke-width:1px
classDef model fill:#201734,stroke:#c6b6ff,color:#f7f3e8,stroke-width:1px
classDef store fill:#142018,stroke:#ddeed3,color:#f7f3e8,stroke-width:1px
classDef output fill:#241515,stroke:#ff9f8a,color:#f7f3e8,stroke-width:1px
A["Desktop Workspace"]:::desktop --> B["Strict Readiness Gate"]:::gate
B --> C["Tavily Research"]:::tool
B --> D["Firecrawl Extraction"]:::tool
B --> E["Composio Tool Graph"]:::tool
B --> F["E2B Sandbox Analysis"]:::tool
C --> G["Evidence Bundle"]:::store
D --> G
E --> G
F --> G
G --> H["Gemini Operating Model"]:::model
H --> I["Artifacts + Report"]:::output
I --> J["Supabase Sync"]:::store
J --> K["Export + Audit Trail"]:::output
| Endpoint | Purpose |
|---|---|
GET /api/health |
Service health and integration readiness |
GET /api/config/status |
Masked .env readiness status |
POST /api/agent/runs |
Execute a connected agent run |
GET /api/agent/runs |
List saved runs |
GET /api/agent/runs/:id/export |
Export JSON and markdown |
GET /api/audit |
Audit history |
- Product website opens at
/. - Desktop command workspace opens at
/app. - Missing
.envintegrations lock the Run button and return503. - Connected runs show Gemini, Tavily, Firecrawl, Composio, E2B, and Supabase evidence.
- Exports include provider status, sources, tool-call evidence, sandbox analysis, and artifacts.
npm run checkpasses before release.
Built as a connected agent product for real company workflows.