Backup Routing

What it does

Backup Routing tells an agent which of its providers to use right now for a job. Send the job kind (chat-model, embeddings, email-send, payments, object-storage, compute-hosting or code-hosting) and the providers you can use, in your own order. You get them back ranked for right now by one published rule on the latest Dependency Pulse reading of each: 1 steady (up, and the state has held 30 minutes or more), 2 recovering (back up less than 30 minutes ago), 3 degraded, 4 maintenance, 5 no current reading, 6 outage. Inside each group your own order is kept, so the tool never overrides your choice between working providers. Each place comes with the reading behind it (state, since when, our answer time, when it was checked) and the whole order is signed, so you can keep it with your logs and check it with the public key below. It is part of the Smart Hub: neutral routing by a published rule, with limits and a signed record.

Who it is for

Agents, and the people who run them, that can use more than one provider for a job and want to pick a working one before a call instead of failing on the first.

Price

Free, no sign-up: 100 calls a day per caller (resets 00:00 UTC). Paid calls are staged on test money and not charging: today a call past 100 in a day is refused until the next day; when the paid tier is switched on, each call past 100 will cost 0.005 USD, paid per call by the agent (x402).

Limits

Reads Dependency Pulse readings only: one vantage point (LLPRO's server in Texas, USA), every 3 minutes, for the provider or cloud region as a whole, not your account, key, region, model or the exact product you use. A provider can work for us and fail for you, and the reverse; place 1 is not a promise. Up to 10 providers per call, from the fixed list for each job kind (call with only a job kind to see them all). Answer time counts only through the Dependency Pulse rule (over 5,000 ms is degraded); providers are never reordered by speed, price or anything outside the rule. It never calls, tests or pays a provider for you and never holds your keys. Nothing you send is stored; your address is held in memory for the daily allowance only (every call counts, including refused ones). Through the A2A door the answer (public readings and your list of provider ids, nothing else) is kept 7 days so you can fetch the task. The gateway keeps call metadata (tool, status, time, a hashed caller id, country, user agent family, referring site) for 90 days for traffic counts and abuse checks, never the request contents.

For developers: try it

POST https://llpro.ai/mcp
{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"backup_route","arguments":{"job":"chat-model","providers":["openai","anthropic","mistral"]}}}
-> ranked: each provider with its place, group (steady, recovering, degraded, maintenance, no-current-reading, outage), the reading behind it and your own place; first_up (place 1 if it is up, else none); the rule and its sha256; and a signed observation

For agents

Tool backup_route on https://llpro.ai/mcp and in the tools list. Signed answers are checked with the public key below (signer llpro1:e42ugoibimuatti572a6eaqcri): an Ed25519 signature over the observation as canonical JSON (sorted keys, no spaces).

Public key (base64): axcJ3vtqAq9q426xEc8btuH2PJfcLSct55zFOfET934=

This tool exists to be found by agents that search MCP registries for failover, fallback, backup provider or routing tools (listed in MCP registries, served by the Campus gateway: MCP, A2A and REST). Call it before a job that can run on more than one provider, or right after a failed call, then use first_up or walk the order. Errors come back as isError with a plain reason (unknown job kind, a provider not read for that job, more than 10 providers, daily allowance used). Readings come from Dependency Pulse (https://llpro.ai/dependency-pulse); how a stamp is checked: https://llpro.ai/stamp

Next

• Connect your agent: Use this and the other LLPRO tools from your own agent.
• See the readings: The signed readings this order is built on, for every service.
• Hand the whole job to the Hub: Tell the Hub the job in plain words; it routes you to the right signed tools and agents.

For agents: this page as Markdown · JSON · llms.txt · MCP endpoint llpro.ai/mcp · agent card