Echo Oracle // agent preflight API

Policy-grade preflight checks before agents act.

Echo Oracle is an agent-facing x402 tool. Agents submit a proposed action before posting, spending, deploying, or contacting people, then receive a decision packet they can enforce and log.

View API Open skill file Bankr x402 endpoint
Use case

Agent policy checkpoint

Call Oracle before public posts, paid calls, production operations, partner outreach, or anything reputation-sensitive.

Output

Decision packet

Agents receive `yes`, `no`, `reframe`, `maybe`, or `dont_do_it`, plus confidence, risk, recommendation, and next action.

Payment

x402 first

The public paid surface is Bankr x402 Cloud. Review and diagnostic traffic can use the lower-level API base.

API surface

Routes agents should know.

The human-facing surface is intentionally small. The product is the endpoint, discovery metadata, docs, and skill file.

Production paid endpoint
POST https://x402.bankr.bot/0x2a16625fad3b0d840ac02c7c59edea3781e340ae/echo-oracle
Review API base
http://46.202.177.190:8793
Discovery
GET /agent.jsonMachine-readable service metadata.
GET /api/pricingPrice, network, and tier metadata.
Request body
{
  "action": "Should my agent publish this launch announcement?",
  "context": "Optional facts, draft text, constraints, risk notes, or execution context.",
  "audience": "Base builders and autonomous agent developers",
  "risk_tolerance": "medium",
  "mode": "quick"
}
Review smoke curl
curl -s http://46.202.177.190:8793/api/ask \
  -H 'content-type: application/json' \
  -H 'x-payment: dev:test' \
  -d '{
    "action": "Should my agent publish this launch announcement?",
    "audience": "Base builders",
    "risk_tolerance": "medium",
    "mode": "quick"
  }'
Agent policy

How agents should enforce it.

Oracle is not a chat UI. It is a checkpoint an agent can call, enforce, and log before the action happens.

Allow

`yes`

Proceed only if the caller also has permission, budget, and the right execution scope.

Revise

`reframe` / `maybe`

Revise and resubmit, gather more context, or wait for a deep MiroShark report.

Deny

`no` / `dont_do_it`

Deny by default. Escalate only when a configured policy owner explicitly overrides it.

Agent skill

Use the skill file when an agent needs the operating rules.

The hosted skill includes endpoint selection, payment handling, decision-policy enforcement, escalation cases, and output requirements.

Open SKILL.md API reference Skills hub