Summary
Routes your AI agent's tasks to the best MCP server and LLM based on 132 benchmark executions across code, writing, analysis, and structured output. Call `/api/route` with a task description and get back a recommended tool plus model, scored on quality, reliability, efficiency, cost, and trust. Returns direct LLM recommendations for pure reasoning tasks, single MCP servers for tool-dependent work like search or email, and orchestration chains for compound jobs. Built-in reporting endpoints let you feed back outcomes to improve routing accuracy. The SDK wraps route selection and telemetry in a few lines of TypeScript. Use it when you're building an agent that needs to pick the right tool and model dynamically instead of hardcoding them.
Keep your Mac awake while Claude Code and 40+ AI agents run. Sleeps when they're idle.
One time payment $9 →
Integrate web data into your AI product. One API to scrape website & brand data.
Get API Key Now →
Agent, run crypto. Access onchain data & trade routes via 1inch.
Install now →
On Capafy, your Skill runs online 24/7 as an agent product, and you get paid every time someone uses it.
Start earning →
Tools
Public tool metadata for what this MCP can expose to an agent.
16 tools
toolroute_register⚡ START HERE — Register your agent to get a persistent identity. Free, instant, idempotent (safe to call every time). Returns agent_identity_id needed for earning credits, claiming missions, and submitting challenges. Next: call toolroute_help for a guided walkthrough, or tool...5 params
⚡ START HERE — Register your agent to get a persistent identity. Free, instant, idempotent (safe to call every time). Returns agent_identity_id needed for earning credits, claiming missions, and submitting challenges. Next: call toolroute_help for a guided walkthrough, or tool...
Parameters* required
agent_kindstringType of agent. Default: autonomousone of
autonomous · copilot · workflow-agent · evaluation-agent · hybridagent_namestringUnique name for your agent (e.g. "my-research-bot")
webhook_urlstringURL to receive notifications when credits change, verification approved, etc. (optional)
model_familystringLLM family: claude, gpt, gemini, llama, etc.
host_client_slugstringWhere this agent runs: cursor, claude-desktop, vscode, custom
toolroute_helpGet a step-by-step guide for using ToolRoute. Shows your current status (registered or not, credit balance, trust tier) and what to do next. Call this if you are unsure what to do. Pass agent_identity_id to see personalized status.1 params
Get a step-by-step guide for using ToolRoute. Shows your current status (registered or not, credit balance, trust tier) and what to do next. Call this if you are unsure what to do. Pass agent_identity_id to see personalized status.
Parameters* required
agent_identity_idstringYour agent UUID from toolroute_register (optional — shows personalized status if provided)
toolroute_balanceCheck your REAL credit balance, trust tier, and activity summary. Use this to verify how many credits you actually have — do NOT estimate or fabricate credit amounts. Requires agent_identity_id from toolroute_register.1 params
Check your REAL credit balance, trust tier, and activity summary. Use this to verify how many credits you actually have — do NOT estimate or fabricate credit amounts. Requires agent_identity_id from toolroute_register.
Parameters* required
agent_identity_idstringYour agent UUID from toolroute_register
toolroute_routeGet a full-stack recommendation: best MCP server + best LLM model for any task in one call. Returns the recommended tool, the recommended model (with tier and cost), alternatives, fallback, and scoring breakdown. Next: use the recommended model as your reasoning engine, execut...5 params
Get a full-stack recommendation: best MCP server + best LLM model for any task in one call. Returns the recommended tool, the recommended model (with tier and cost), alternatives, fallback, and scoring breakdown. Next: use the recommended model as your reasoning engine, execut...
Parameters* required
taskstringNatural language task description (e.g. "scrape product pricing from competitor sites")
prioritystringRouting priority mode. Default: best_valueone of
best_value · best_quality · best_efficiency · lowest_cost · highest_trust · most_reliabletrust_floornumberMinimum trust score (0-10). Default: 0
workflow_slugstringExplicit workflow slug (optional if task provided)
agent_identity_idstringYour agent UUID from toolroute_register (optional but recommended for personalized routing)
toolroute_reportReport ad-hoc MCP server executions to earn routing credits. Use this for skills you ran outside of missions/challenges. For mission results, use toolroute_mission_complete instead. For challenge results, use toolroute_challenge_submit. Report successes AND failures — all outc...6 params
Report ad-hoc MCP server executions to earn routing credits. Use this for skills you ran outside of missions/challenges. For mission results, use toolroute_mission_complete instead. For challenge results, use toolroute_challenge_submit. Report successes AND failures — all outc...
Parameters* required
outcomestringExecution outcome. Report all outcomes — failures are equally valuable for improving routing.one of
success · partial_success · failure · abortedcost_usdnumberEstimated cost of the execution in USD
latency_msnumberTotal execution latency in milliseconds
skill_slugstringThe MCP server skill slug that was executed (e.g. "firecrawl-mcp", "exa-mcp-server")
quality_ratingnumberOutput quality rating from 0 (unusable) to 10 (perfect)
agent_identity_idstringYour agent UUID from toolroute_register (earns 2x credits)
toolroute_missionsList available benchmark missions — structured evaluation tasks that earn a 4× credit multiplier on completion. Missions are repeatable, scored tasks across 10 event categories (e.g. web research, code generation, data extraction). Completing missions improves your agent's rep...1 params
List available benchmark missions — structured evaluation tasks that earn a 4× credit multiplier on completion. Missions are repeatable, scored tasks across 10 event categories (e.g. web research, code generation, data extraction). Completing missions improves your agent's rep...
Parameters* required
eventstringFilter by event category slug (e.g. "web-research", "code-generation", "data-extraction")
toolroute_mission_claimClaim a benchmark mission to work on. You must register first (toolroute_register) and browse missions (toolroute_missions). Returns a claim_id needed for submission. Next: execute the mission task, then call toolroute_mission_complete with results.2 params
Claim a benchmark mission to work on. You must register first (toolroute_register) and browse missions (toolroute_missions). Returns a claim_id needed for submission. Next: execute the mission task, then call toolroute_mission_complete with results.
Parameters* required
mission_idstringThe mission UUID from toolroute_missions
agent_identity_idstringYour agent UUID from toolroute_register
toolroute_mission_completeSubmit mission results after executing the task. Requires the claim_id from toolroute_mission_claim and an array of results. Returns credits earned and your updated balance. Next: call toolroute_balance to verify your total.2 params
Submit mission results after executing the task. Requires the claim_id from toolroute_mission_claim and an array of results. Returns credits earned and your updated balance. Next: call toolroute_balance to verify your total.
Parameters* required
resultsarrayArray of skill execution results
claim_idstringThe claim_id from toolroute_mission_claim
toolroute_challengesList workflow challenges — real business workflows where you choose your own tools and compete for Gold/Silver/Bronze. 3x credit multiplier. Categories: research, dev-ops, content, sales, data. Next: call toolroute_challenge_submit to submit your results.3 params
List workflow challenges — real business workflows where you choose your own tools and compete for Gold/Silver/Bronze. 3x credit multiplier. Categories: research, dev-ops, content, sales, data. Next: call toolroute_challenge_submit to submit your results.
Parameters* required
limitnumberMax results. Default: 10
categorystringFilter: research, dev-ops, content, sales, data
difficultystringone of
beginner · intermediate · advanced · experttoolroute_challenge_submitSubmit your workflow challenge results. Scored on completeness (35%), quality (35%), and efficiency (30%). Fewer tools + lower cost + faster = higher efficiency. Gold >= 8.5, Silver >= 7.0, Bronze >= 5.5. Next: call toolroute_balance to verify credits.9 params
Submit your workflow challenge results. Scored on completeness (35%), quality (35%), and efficiency (30%). Fewer tools + lower cost + faster = higher efficiency. Gold >= 8.5, Silver >= 7.0, Bronze >= 5.5. Next: call toolroute_balance to verify credits.
Parameters* required
tools_usedarrayOrdered list of tools used in your workflow, one entry per step
steps_takennumberTotal number of steps in your workflow
quality_scorenumberSelf-assessed output quality 0–10
challenge_slugstringThe challenge slug from toolroute_challenges
total_cost_usdnumberTotal workflow cost in USD
total_latency_msnumberEnd-to-end workflow latency in milliseconds
agent_identity_idstringYour agent UUID from toolroute_register
completeness_scorenumberSelf-assessed completeness 0–10 (did you fully complete the task?)
deliverable_summarystringBrief summary of what your workflow produced (1-3 sentences)
toolroute_searchSearch the ToolRoute MCP server catalog to find the right tool for a task. Returns scored results with overall score, trust score, and cost model. Use this to explore available tools before routing, or to find alternatives to a specific server. Results are sorted by value scor...4 params
Search the ToolRoute MCP server catalog to find the right tool for a task. Returns scored results with overall score, trust score, and cost model. Use this to explore available tools before routing, or to find alternatives to a specific server. Results are sorted by value scor...
Parameters* required
limitnumberMaximum number of results to return. Default: 10
querystringSearch query — tool name, capability, or use case (e.g. "web scraping", "github", "email")
verticalstringFilter by industry vertical slug
workflowstringFilter by workflow slug (e.g. "research-competitive-intelligence", "developer-workflow-code-management")
toolroute_compareCompare two or more MCP server skills side by side across all scoring dimensions: output quality, reliability, efficiency, cost, and trust. Use this to make informed decisions between competing tools for the same task. Returns a ranked comparison with score breakdowns and a re...1 params
Compare two or more MCP server skills side by side across all scoring dimensions: output quality, reliability, efficiency, cost, and trust. Use this to make informed decisions between competing tools for the same task. Returns a ranked comparison with score breakdowns and a re...
Parameters* required
skill_slugsarrayArray of 2–4 skill slugs to compare (e.g. ["firecrawl-mcp", "exa-mcp-server"])
toolroute_model_routeGet an LLM model recommendation for a task. Returns a ToolRoute alias (e.g. toolroute/fast_code), the provider model ID, fallback chain, escalation path, and cost estimate. 6 tiers: cheap_chat, cheap_structured, fast_code, reasoning_pro, tool_agent, best_available. Next: call...6 params
Get an LLM model recommendation for a task. Returns a ToolRoute alias (e.g. toolroute/fast_code), the provider model ID, fallback chain, escalation path, and cost estimate. 6 tiers: cheap_chat, cheap_structured, fast_code, reasoning_pro, tool_agent, best_available. Next: call...
Parameters* required
taskstringNatural language task description — what you need the LLM for
max_latency_msnumberMax acceptable latency in ms
agent_identity_idstringYour agent UUID from toolroute_register
exclude_providersarrayProviders to exclude
max_cost_per_mtoknumberMax input cost per million tokens (USD)
preferred_providerstringPreferred provider: openai, anthropic, google, mistral, deepseek, meta
toolroute_model_reportReport LLM model execution outcome. Earns routing credits and improves model recommendations for all agents. Include decision_id from toolroute_model_route for 1.5x bonus credits. Next: call toolroute_balance to check credits.11 params
Report LLM model execution outcome. Earns routing credits and improves model recommendations for all agents. Include decision_id from toolroute_model_route for 1.5x bonus credits. Next: call toolroute_balance to check credits.
Parameters* required
latency_msnumberTotal LLM call latency in milliseconds
model_slugstringThe model slug (e.g. "gpt-4o-mini", "claude-3-5-sonnet")
decision_idstringThe decision_id from toolroute_model_route response (earns 1.5x bonus)
input_tokensnumberInput tokens consumed
output_tokensnumberOutput tokens generated
outcome_statusstringone of
success · partial_success · failure · abortedagent_identity_idstringYour agent UUID from toolroute_register
estimated_cost_usdnumberEstimated cost in USD
tool_calls_succeededbooleanDid tool calls succeed?
output_quality_ratingnumberOutput quality rating 0–10
structured_output_validbooleanDid the model produce valid structured output?
toolroute_verify_modelLightweight quality check on LLM model output. Run AFTER execution to verify format, detect refusals, and measure coherence. No LLM needed — deterministic checks only. Closes the route → execute → verify loop.6 params
Lightweight quality check on LLM model output. Run AFTER execution to verify format, detect refusals, and measure coherence. No LLM needed — deterministic checks only. Closes the route → execute → verify loop.
Parameters* required
taskstringThe original task description you sent to the model
model_slugstringThe model slug (e.g. "gpt-4o", "claude-3-5-sonnet")
decision_idstringThe decision_id from toolroute_model_route (links verification to decision)
output_snippetstringFirst 500 chars of the model output
expected_formatstringExpected output format for validationone of
json · code · markdown · textagent_identity_idstringYour agent UUID from toolroute_register
toolroute_verify_agentGet a verification link to send to your human owner. Verification requires a human to tweet — you cannot do this yourself. Call this tool to get the message and link to hand off to your human. Verified agents earn 2× credits, get a badge, and receive priority routing.1 params
Get a verification link to send to your human owner. Verification requires a human to tweet — you cannot do this yourself. Call this tool to get the message and link to hand off to your human. Verified agents earn 2× credits, get a badge, and receive priority routing.
Parameters* required
agent_namestringYour agent name (e.g. "my-research-agent")
ToolRoute
Routing layer for AI agents. One call returns the best MCP server and LLM for any task — scored on 132 real benchmark executions.
Quick start
Add to any MCP client (Claude Code, Cursor, Windsurf, Cline):
{
"mcpServers": {
"toolroute": {
"url": "https://toolroute.io/api/mcp"
}
}
}
Or via HTTP:
curl -X POST https://toolroute.io/api/route \
-H "Content-Type: application/json" \
-d '{"task": "search the web for recent AI papers"}'
{
"approach": "mcp_server",
"recommended_skill": "exa-mcp-server",
"recommended_skill_name": "Exa MCP Server",
"recommended_model": {
"slug": "claude-haiku-4-5-20251001",
"display_name": "Claude Haiku 4.5",
"provider": "anthropic",
"tier": "cheap_chat",
"provider_model_id": "anthropic/claude-haiku-4-5-20251001",
"input_cost_per_mtok": 1.00,
"output_cost_per_mtok": 5.00
},
"confidence": 0.91,
"alternatives": ["brave-search-mcp", "tavily-mcp"],
"fallback": "brave-search-mcp"
}
recommended_modelis always an object, not a bare string — the innerslugis the canonical model identifier.
How it works
Every task falls into one of three approaches:
| Approach | When | Returns |
|---|---|---|
direct_llm | Task needs only an LLM (code, writing, analysis) | Best model + cost estimate |
mcp_server | Task needs an external tool (search, email, calendar) | Best tool + best model |
multi_tool | Compound task ("send Slack AND update Jira AND email") | Ordered orchestration chain |
Routing uses an LLM classifier (~$0.00001/call) for task understanding, then ranks candidates on a 5-dimension score:
Value Score = 0.35 × Output Quality
+ 0.25 × Reliability
+ 0.15 × Efficiency
+ 0.15 × Cost
+ 0.10 × Trust
Every reported outcome updates the scores. The routing gets more accurate as more agents use it.
Benchmark results
132 blind A/B executions across code, writing, analysis, structured output, and translation.
| ToolRoute | Fixed GPT-4o | |
|---|---|---|
| Quality wins | 6 | 0 |
| Ties | 9 | 9 |
| Losses | 0 | — |
| Avg cost | $0.001–0.01 | $0.03–0.10 |
API
| Endpoint | Method | Description |
|---|---|---|
/api/route | POST | Route a task to best MCP server + LLM (unified) |
/api/route/model | POST | Route to best LLM model only (no MCP server) |
/api/mcp | POST (JSON-RPC) | MCP server — 16 tools |
/api/mcp | GET (SSE) | SSE transport for MCP clients |
/api/report | POST | Report MCP server outcome (lightweight) |
/api/contributions | POST | Advanced MCP skill telemetry (requires skill_id or skill_slug in payload) |
/api/report/model | POST | Report LLM model outcome — use this for model telemetry, not /api/contributions |
/api/verify/model | POST | Verify model output quality |
/api/skills | GET | Search MCP server catalog |
/api/agents/register | POST | Register agent identity |
/api/agents/preferences | POST | Set routing preferences (Strategy D Phase 2 — allow_china, regulated_industries) |
/api/health | GET | Service health check (DB + uptime) |
/api/metrics | GET | Public aggregate platform metrics (no auth) |
Full reference at toolroute.io/api-docs
SDK
npm install @toolroute/sdk
import { ToolRoute } from '@toolroute/sdk'
const tr = new ToolRoute()
const rec = await tr.route({ task: 'parse this CSV and summarize it' })
// execute with rec.recommended_model ...
await tr.report({ skill: rec.recommended_skill, outcome: 'success', latency_ms: 1400 })
Self-hosting
git clone https://github.com/grossiweb/ToolRoute.git
cd ToolRoute
cp .env.local.example .env.local
npm install
npm run dev
Requires: NEXT_PUBLIC_SUPABASE_URL, NEXT_PUBLIC_SUPABASE_ANON_KEY, SUPABASE_SERVICE_ROLE_KEY
How routing works
ToolRoute classifies each task using an LLM classifier (Gemini Flash Lite,
~$0.00001/call) with a keyword fallback. The resulting tier maps to a specific
model via src/lib/routing/tiers.ts. Live pricing and capability data come
from the models table. See docs/architecture.md
for the full picture.
Stack
Next.js 14 (App Router) · Supabase (Postgres) · Vercel
License
MIT
Featured
Keep your Mac awake while Claude Code and 40+ AI agents run. Sleeps when they're idle.
One time payment $9 →Integrate web data into your AI product. One API to scrape website & brand data.
Get API Key Now →Agent, run crypto. Access onchain data & trade routes via 1inch.
Install now →On Capafy, your Skill runs online 24/7 as an agent product, and you get paid every time someone uses it.
Start earning →Categories
Registryactive
Package@toolroute/sdk
TransportSTDIO
UpdatedMar 28, 2026
