Sato Bot
Describe the agent, get the build plan
Say what you want the agent to do. Sato Bot answers with a stack of real listings — each carrying its Sato Score, its Preflight verdict and its install line — the first action your goal implies, and the questions you still have to answer. No model writes the names: every component is a directory record, and unknown stays unknown.
A plan is a recommendation, not an assurance
- Every stack item is a directory listing; nothing in a plan is invented.
- A Sato Score measures how open, active and verifiable a project is — never a security, quality or returns judgment.
- A Preflight verdict names what was checked and when. `unknown` means Sato Hub holds no record, not that something is wrong.
- A route is a recommendation chosen on the fields named in chosen_by; a quote is a quote, not a fill.
- Nothing is signed, deployed or broadcast here. The caller holds its own keys.
- null is unknown and is never read as zero.
Try it
This form calls the same endpoint an agent calls, and prints what came back — nothing is added or softened in between.
What a plan contains
- The stack
- Real directory listings, one per slot, each with its Sato Score, its liveness, the observed success rate of Sato Hub's own daily checks, and its canonical page. A project that is not in the directory cannot appear in a plan — there is no model here writing plausible names.
- A Preflight verdict per component
- go, caution, no or unknown for the top items, each naming the rule that decided it and the date the fields were written. unknown means Sato Hub holds no record, which is an absence rather than a warning.
- The install line
- For every component that publishes a deploy spec: the runtime, the first install command and what it requires before it will run. Where the install has been reproduced in a container, the plan says so.
- The first action
- When the goal implies one transaction — a swap, a token launch — the plan carries a route for it, with the fee disclosed. A swap is quoted at a NOMINAL size, never a size we assumed for you, and a launch config is prepared for you to read and sign yourself.
- The questions back
- Who holds the keys, what the agent may sign without you, what triggers an action and what happens when it is wrong. A plan that does not ask these is not a plan.
Call it
HTTP
curl -s -X POST "https://satohub.ai/api/satobot/plan" \
-H 'content-type: application/json' \
-d '{"goal":"a Base trading agent that swaps USDC to ETH on a signal","chain":"Base"}'GET works too, with the same parameters in the query string. Add format=markdown for a chat-shaped render.
MCP
{
"mcpServers": {
"satohub": { "url": "https://satohub.ai/api/mcp" }
}
}The tool is onchain_agent_build_plan. Full list on /mcp.
Telegram: an adapter runs on the same planner and is coming — not open to the public yet. The page and the API are the two ways in today.
Questions
What is Sato Bot?
A planner. You describe the onchain agent you want in plain words, and it returns a build plan: a stack of real Sato Hub directory listings, the deploy spec for each item that publishes one, a Preflight verdict per item, the first action when the goal implies one, the questions you still have to answer, and the next steps.
Is there a model writing these plans?
No. Intent and chain are parsed by small rule tables, and every name, number, score and URL comes from Sato Hub's own records. A listing that is not in the directory cannot appear in a plan, and unknown stays unknown rather than being filled in with a plausible value.
Does Sato Bot hold keys or sign anything?
No. It holds no keys, signs nothing, deploys nothing and broadcasts nothing. A swap first-action is a quote taken at a nominal size; a launch first-action is a config you read and sign yourself, with the fee disclosed before anything is signed. An unsigned transaction pays nothing.
Does appearing in a plan vouch for a component?
No. A Sato Score measures how open, active and verifiable a project is — it says nothing about security, quality or returns. A Preflight verdict describes what was checked and when. A plan is a recommendation assembled from dated records, not a security review.
Can my own agent call Sato Bot?
Yes. POST https://satohub.ai/api/satobot/plan with a goal, or call the tool onchain_agent_build_plan over MCP at https://satohub.ai/api/mcp. Both run the same brain, so the web page and an agent's tool loop cannot be given different answers.
Is Sato Bot on Telegram?
A Telegram adapter is built against the same planner and is coming; it is not open to the public yet. Until it is, the page here and the API are the two ways in.
Sato Bot reports readings Sato Hub holds. It is not a security audit, a code review or financial advice. Related: Preflight · Sato Route · Sato Score · Directory