ToolYour
Remote MCP: plan, run, verify for SEO/security/ship-gate. Same REST API key; 500 free credits/month.
التثبيت
أضف هذا إلى إعدادات MCP في عميلك (Claude Desktop: claude_desktop_config.json، وCursor: .cursor/mcp.json، ويستخدم VS Code وغيره الصيغة نفسها «mcpServers»).
claude mcp add --transport http toolyour 'https://api.toolyour.com/mcp/http' --header 'X-Api-Key: <ToolYour API key from the dashboard (prefix ty_). Same key as REST.>'
{
"mcpServers": {
"toolyour": {
"type": "http",
"url": "https://api.toolyour.com/mcp/http",
"headers": {
"X-Api-Key": "<ToolYour API key from the dashboard (prefix ty_). Same key as REST.>"
}
}
}
}
claude mcp add --transport sse toolyour 'https://api.toolyour.com/mcp' --header 'X-Api-Key: <ToolYour API key from the dashboard (prefix ty_). Same key as REST.>'
{
"mcpServers": {
"toolyour": {
"type": "sse",
"url": "https://api.toolyour.com/mcp",
"headers": {
"X-Api-Key": "<ToolYour API key from the dashboard (prefix ty_). Same key as REST.>"
}
}
}
}
التوثيق
مأخوذ من ملف README للمشروع على GitHub ويُحدَّث أسبوعيًا. تفتح الروابط على GitHub.
المحتويات
Open-source ToolYour MCP
Remote MCP gateway for AI agents: plan → run playbook/solve → verify until pass.
MIT licensed. Hosted API/SaaS backends are optional — offline evals and local file control-plane run without them.
Product docs: https://www.toolyour.com/developers/mcp
Official registry: com.toolyour/mcp
npm client (separate): @toolyour/sdk (MIT)
Quick start
cp .env.example .env
npm install
npm run build:registry # from monorepo docs/ when regenerating
npm run build
npm run dev
Endpoints: http://localhost:3090/mcp (legacy SSE for compatible hosts) · Streamable HTTP: /mcp or /mcp/http.
The API-key Streamable HTTP endpoints are stateless: clients do not need session
affinity, and a stale Mcp-Session-Id is ignored. The OAuth-protected
/mcp/chatgpt endpoint has its own session behavior.
Health: GET /health/mcp · Ready: GET /health/mcp/ready · Metrics: GET /health/mcp/metrics (Prometheus text)
Auth
X-Api-Key: ty_...
Same key as ToolYour REST. Free tier: 500 credits/month (tools cost 1–10 credits). MCP exposes API-backed catalog tools only (hasApi).
Host contract
ToolYour does not replace Cursor/Claude. The host keeps editor, git, and terminal. Agents should:
plan_task(free)run_playbookorsolve_task- Apply only rank-1
loop.nextActionsin the workspace verify_taskwith prior result as baseline untilloop.gateispass(or stop)
Skill-loop responses also include an additive execution envelope (runId, intentId,
intentType, idempotencyKey, and projectScope). Preserve it when handing a result to
another MCP tool. Async jobs reuse execution.runId as the get_run polling id.
Internally, plan_task, solve_task, run_playbook, run_workflow, and verify_task use
one executeIntent lifecycle boundary for correlation and the execution envelope.
Credits settle on each gateway tool invoke (billing.settledBy = gateway_per_tool) — MCP
does not double-bill. Free recall_context recalls purpose-typed memory before rebuilds.
Saved Playbooks (private beta)
Saved Playbooks turns a repeatable ToolYour workflow into a named, versioned recipe that an MCP-capable agent can run again. V1 deliberately points only to existing ToolYour skills or workflows; it does not store arbitrary shell commands, control another agent, or bypass host approval. Each account gets owner-scoped playbooks and durable run receipts.
Enable this only on a private/default MCP deployment with PLAYBOOKS_BETA=true. The beta adds
create_saved_playbook, list_saved_playbooks, get_saved_playbook, update_saved_playbook,
run_saved_playbook, and get_saved_playbook_run. It is intentionally excluded from
/mcp/chatgpt so the reviewed public 21-tool surface stays unchanged.
See Saved Playbooks beta design for the data model, safe execution boundary, and rollout guidance.
Offline evals (no API key)
npm run ci # build + lint + unit/contract/integration + evals
npm run eval:golden # routing + jobReport fields + verify gate
npm run eval:playbooks # skill → workflow → synthesizer matrix + fixture coverage
npm run golden:tier1 # host work-package shapes
Fixtures: tests/eval/synth-fixtures.json · Goals: tests/eval/goals.jsonl
Observability (open protocols)
| Signal | How |
|---|---|
| JSON logs | Always (LOG_LEVEL) |
| In-process counters | /health/mcp + /health/mcp/metrics |
| Spans | Set OTEL_LOG_SPANS=1 and/or OTEL_EXPORTER_OTLP_ENDPOINT (OTLP/HTTP JSON → Collector / Jaeger / Tempo) |
No proprietary APM required.
What stays hosted (optional)
- Live tool execution via gateway + API key
- Feature Memory / verification profiles (SaaS Mongo)
- Redis for multi-replica async
get_run
Local defaults use file-backed control-plane jobs.
License
MIT — see LICENSE.
الأسئلة الشائعة
ما هو ToolYour؟
Remote MCP: plan, run, verify for SEO/security/ship-gate. Same REST API key; 500 free credits/month.
كيف أثبّت ToolYour؟
يعمل كخادم بعيد. في Claude Code: claude mcp add --transport http toolyour 'https://api.toolyour.com/mcp/http' --header 'X-Api-Key: <ToolYour API key from the dashboard (prefix ty_). Same key as REST.>'. وفي العملاء الآخرين، الصق إعداد JSON من هذه الصفحة في إعدادات MCP الخاصة بهم.
هل يحتاج ToolYour إلى مفتاح API؟
نعم. يتطلب X-Api-Key، وتضبطها في إعداداته.
هل ToolYour مفتوح المصدر؟
نعم. شيفرته متاحة على GitHub بترخيص MIT.
من يصون ToolYour؟
ينشره AbdulWahabRaza123. يدرجه AI.tn لكنه لا يصونه ولا يوصي به.
من سجل MCP الرسمي. الإدراج ليس توصية: راجع شيفرة الخادم وصلاحياته قبل منحه الوصول إلى بياناتك.