Remote MCP server for Cursor and Claude

Why do AI agents need an MCP harness?

Most MCP servers cover code, repos, or browsers. ToolYour is the plan → run → verify utilities layer — ship gates, SEO audits, security checks, and related jobs from Cursor or Claude without per-tool wrappers. It does not replace your agent runtime.

• Task-oriented

  Start with plan_task (free), then run_playbook or solve_task. Read loop.remainingFixes, apply them in the repo, then verify_task until loop.gate is pass. invoke_tool is only for a one-off operationId.

• Token-efficient discovery

  Agents search the catalog on demand instead of loading hundreds of tool schemas upfront.

• Same plan as REST

  One key, one quota for apps and agents. Free tier includes 500 credits/month.

Which job workflows does solve_task route to?

solve_task is the primary MCP entry: plain-language goals route to workflows or tools with fuzzy matching, confidence gating, and synthesized job reports. Ambiguous goals return ranked suggestions instead of a wrong run. When a goal matches, workflows return a merged jobReport with scores, findings, and prioritized actions. Discovery and suggestions without execution remain free.

• full-seo-audit

  SEO + page speed audit

• core-web-vitals-job

  Core Web Vitals diagnosis

• full-seo-optimization-job

  Full SEO optimization (6 tools)

• internal-link-architecture-job

  Internal link architecture

• technical-seo-audit-job

  Technical SEO audit (lite)

• social-preview-audit-job

  Social preview (OG / Twitter)

• content-quality-audit-job

  Content quality audit

• keyword-opportunity-review-job

  Keyword opportunity review

• seo-deploy-regression-job

  SEO deploy regression (bulk URLs)

• seo-deploy-regression-diff-job

  SEO staging vs production diff

• crawl-readiness-job

  Crawl readiness (robots, sitemap, redirects)

• structured-data-audit-job

  Structured data audit (schema + meta)

• ai-overview-readiness-job

  AI Overview readiness check

• site-icons-audit-job

  Site icons audit (favicon, apple-touch)

• document-convert-pipeline

  Document convert pipeline

• full-security-audit

  Full security audit (headers + TLS + cookies)

• security-headers-job

  Security headers check

• frontend-supply-chain-job

  Frontend supply chain

• developer-ship-checklist-job

  Developer ship checklist

• dev-json-pipeline-job

  Dev JSON pipeline

• dev-auth-debug-job

  Dev auth debug

• dev-format-transform-job

  Dev format transform

• ship-gate-job

  Ship gate (pass/fail deploy)

• secrets-hygiene-job

  Secrets hygiene scan

• auth-token-hygiene-job

  Auth token hygiene

• email-auth-security-job

  Email auth + DNS + security.txt

• campaign-tracking-setup-job

  Campaign tracking setup

• paid-ads-copy-gate-job

  Paid ads copy gate

• growth-unit-economics-job

  Growth unit economics

• email-campaign-qa-job

  Email campaign QA

• landing-conversion-check-job

  Landing conversion check

Curated skill → workflow map: MCP playbooks. Canonical agent loop: plan_task → run_playbook / solve_task → apply loop.remainingFixes → verify_task.

How do you connect Cursor, Claude, or another MCP client?

1. 1

   Get an API key

   Create a key on your dashboard. MCP shares the same monthly quota as REST.

2. 2

   Add the remote MCP URL

   Point your client at https://api.toolyour.com/mcp with header X-Api-Key. Use the config panel above for Cursor (~/.cursor/mcp.json), Claude Desktop, Claude Code, and other MCP-compatible agents.

3. 3

   Agent workflow

   Try plan_task then run_playbook or solve_task — describe the job in plain language (e.g. "ship gate for https://preview.example.com"). solve_task is the primary MCP entry: plain-language goals route to workflows or tools with fuzzy matching, confidence gating, and synthesized job reports. Ambiguous goals return ranked suggestions instead of a wrong run. Ambiguous goals return status: suggest with ranked taskSuggestions.

   1. plan_task (free) then run_playbook or solve_task
   2. Apply loop.remainingFixes in the host repo (editor/git)
   3. verify_task with that result as baseline until loop.gate is pass

   Discovery meta-tools are free and do not count toward quota. Execution bills like REST.

How do browser, REST, and MCP share one ToolYour catalog?

Every API-backed tool is available free in the browser, via REST for apps, and via MCP for agents. Browser-only tools stay on the website. This page is setup, not a ranking list — the utilities-MCP comparison article covers that.

Guides: Best utilities MCP servers · Claude Code MCP setup · invoke_tool vs solve_task.

What are common ToolYour MCP questions?

How does this compare to Zapier MCP or GitHub MCP?

GitHub and Playwright MCPs serve repos and browsers. Zapier MCP connects agents to other apps. ToolYour is a hosted utilities harness (SEO, ship-gate, security, converters) with the same key as REST. Install those servers too when you need their jobs.

Is this a local or remote MCP server?

Remote. Agents connect over HTTPS to https://api.toolyour.com/mcp — no local install of ToolYour tools required.

MCP vs REST — when should I use which?

Use REST when your app or backend calls tools programmatically. Use MCP when an AI agent (Cursor, Claude, etc.) should discover and run tools in conversation. Both share billing.

How mature is solve_task?

solve_task is the primary MCP entry: plain-language goals route to workflows or tools with fuzzy matching, confidence gating, and synthesized job reports. Ambiguous goals return ranked suggestions instead of a wrong run. Prefer plan_task (free) then run_playbook or solve_task(compact). Read loop.remainingFixes, apply them in the host repo, then verify_task with the prior result as baseline until loop.gate is pass. Use discover_tools → invoke_tool only for an explicit one-off operationId, not for ship/SEO/security jobs.

What categories work over MCP?

File converters, SEO & site tools, documents, text utilities, unit converters, finance calculators, and AI utilities — any tool with API access enabled.

Full reference: MCP quickstart docs.