AI tools that run free in your browser. No cloud, no API keys, no subscription. Models run on your hardware via WebGPU/WASM.
Live: https://freeagentstore.online
MCP: npx mcp-remote https://mcp.freeagentstore.online/mcp
GitHub: https://github.com/FreeAgentStore
git clone https://github.com/FreeAgentStore/platform.git
cd platform
pnpm install
pnpm build
pnpm test # 107 tests
pnpm lint # biomeplatform/
├── packages/
│ ├── sdk/ @freeagentstore/sdk — model loading, inference, Ollama, heuristics
│ ├── cli/ @freeagentstore/cli — `fags` binary (init, check, publish)
│ ├── compliance/ Agent-specific compliance checks (9 checks)
│ ├── backend/ API Worker (placeholder)
│ └── quality/ Quality reporter (placeholder)
├── workers/
│ ├── host/ *.freeagentstore.online → D1 → R2 host worker
│ └── mcp/ MCP server (11 tools) at mcp.freeagentstore.online
├── agents/
│ ├── tts/ Text to Speech (Kokoro 82M)
│ ├── transcriber/ Speech to Text (Whisper Small)
│ ├── bg-remover/ Background Removal (RMBG 1.4)
│ └── sentiment/ Sentiment Analysis (heuristic — no model)
├── templates/
│ └── template-agent-tts/ Scaffold for new TTS agents
└── store/
├── index.html Store landing page
└── registry.json Agent catalog (4 agents)
| Type | Runtime cost | Example |
|---|---|---|
| Model-based | Zero (user's GPU/CPU) | TTS, Transcriber, BG Remover |
| Heuristic | Zero (pure JS) | Sentiment Analyzer |
Model-based agents download an AI model once (~150-250MB), cache it in Cache Storage, and run inference in a Web Worker via WebGPU or WASM fallback.
Heuristic agents are pure JS functions generated by an LLM and evolved through iterative testing. No model, no download, instant results.
import { initAgent } from '@freeagentstore/sdk';
import { useModel, useWorkerInference, useOllama } from '@freeagentstore/sdk/hooks';
const agent = initAgent({ agentId: 'my-agent' });
// Model loading with WebGPU/WASM fallback
const { status, load } = useModel({ repo: 'onnx-community/whisper-small', device: 'auto' });
// Web Worker inference (off main thread)
const { run, result } = useWorkerInference({ workerUrl: '/worker.js' });
// Optional local LLM via Ollama
const { available, chat } = useOllama();
// Heuristic agents (no model needed)
import { evaluateHeuristic, buildEvolvePrompt } from '@freeagentstore/sdk';Connect via npx mcp-remote https://mcp.freeagentstore.online/mcp
| Tool | Auth | What it does |
|---|---|---|
list_agents |
Optional | List all or your agents |
agent_info |
No | Status, URLs, links |
deploy_status |
No | GitHub Actions history |
create_agent |
Yes | Provision repo + R2 route |
delete_agent |
Yes | Remove from store |
write_file |
Yes | Commit to agent repo |
read_file |
No | Read from agent repo |
list_files |
No | Directory listing |
upload_to_r2 |
Yes | Trigger redeploy |
platform_guide |
No | Architecture guide |
sdk_reference |
No | API reference |
| Resource | Name/ID |
|---|---|
| CF Zone | freeagentstore.online |
| D1 | fags |
| R2 | fags-agents |
| KV (OAuth) | FAGS_OAUTH_KV |
| Doppler | project fags |
| GitHub Org | FreeAgentStore |
| Workflow | Trigger | What it does |
|---|---|---|
ci.yml |
Push/PR to main | Build + lint + 107 tests |
publish.yml |
Push to packages/ | npm publish with OIDC provenance |
deploy-host.yml |
Push to workers/host/ | D1 migration + wrangler deploy |
deploy-mcp.yml |
Push to workers/mcp/ | wrangler deploy |
| Store | What it hosts | URL |
|---|---|---|
| FreeAppStore | PWA apps | https://freeappstore.online |
| FreeGameStore | PWA games | https://freegamestore.online |
| FreeWebStore | One-page sites | https://freewebstore.online |
| FreeAgentStore | Browser AI tools | https://freeagentstore.online |
| ProAgentStore | Server AI agents | https://proagentstore.online |
MIT