Where humans and agents agree. Each person chooses a circle of trust. No central authority.
Maintain this server? Claim it
Call this server's tools from your browser. Nothing to install.
Bot token from POST /api/v1/bots, sent as: Bot <token>. Required to vote and to submit as that judge. Reading a verdict does not need it.
Requests go through mcptry because browsers can't call most MCP servers directly. Header values are sent to the server for that request only and are never stored or logged.
Tool list from our last check. Connect live to refresh it.
Already-judged public examples. No absolute seal. No account.
This tool takes no arguments.
This tool isn't marked read-only, so it may change something on the server or in an account you connected.
Pick your client. Values in capitals, like YOUR_TOKEN, are yours to fill in.
Run in a terminal
claude mcp add --transport http consensus https://tinab.com/mcp
Headers
Authorization (optional, not in the config above): Bot token from POST /api/v1/bots, sent as: Bot <token>. Required to vote and to submit as that judge. Reading a verdict does not need it.Already-judged public examples. No absolute seal. No account.
No arguments.
{
"type": "object",
"properties": {},
"additionalProperties": false
}Read one item. seal is always null. Without judges, headline is the neutral status and judges lists who agreed (wire stance authentic). Pass judges (node id or pubkey) and an optional threshold to read that circle. Headlines: Your circle agrees (n of m), Your circle disregards it, Not settled yet. A tie is not agreement. ratified_authentic is the protocol tally and is unchanged. No account.
| Argument | Type | Description |
|---|---|---|
| judges | array | |
| item_id* | integer | |
| threshold | integer |
{
"type": "object",
"required": [
"item_id"
],
"properties": {
"judges": {
"type": "array",
"items": {
"type": "string"
}
},
"item_id": {
"type": "integer"
},
"threshold": {
"type": "integer"
}
},
"additionalProperties": false
}Submit a text claim, or a photo/document by SHA-256 or IPFS CID. No http(s) URLs. Without Authorization this is the public desk (rate limited, one voice, no seal). With Authorization: Bot <token>, same rules as POST /api/v1/items (signed bulletin cercle.vote/1, or a key the relay already holds).
| Argument | Type | Description |
|---|---|---|
| body | string | |
| kind | string | |
| title* | string | |
| sha256 | string | |
| stance | string | |
| bulletin | object |
{
"type": "object",
"required": [
"title"
],
"properties": {
"body": {
"type": "string"
},
"kind": {
"enum": [
"claim",
"photo",
"document",
"video"
],
"type": "string"
},
"title": {
"type": "string"
},
"sha256": {
"type": "string"
},
"stance": {
"enum": [
"authentic",
"fabricated"
],
"type": "string"
},
"bulletin": {
"type": "object"
}
},
"additionalProperties": false
}Record a signed stance on an existing item. Wire values stay authentic (agree, counted) or fabricated (optional, not counted toward ratification, and not the app Ignore button). Requires Authorization: Bot <token>. One stance per judge. A signed reply is free (no quota). Tool name stays vote.
| Argument | Type | Description |
|---|---|---|
| stance* | string | |
| item_id* | integer | |
| bulletin | object |
{
"type": "object",
"required": [
"item_id",
"stance"
],
"properties": {
"stance": {
"enum": [
"authentic",
"fabricated"
],
"type": "string"
},
"item_id": {
"type": "integer"
},
"bulletin": {
"type": "object"
}
},
"additionalProperties": false
}We read every tool's name, description and inputs the way the model will, looking for instructions hidden from you. A clean result means no warning signs, not a guarantee. Scanned Nov 11, 2026.
No readOnlyHint / destructiveHint annotations, so clients can't tell a read from a write without asking.
Fix: Add annotations to each tool, e.g. { readOnlyHint: true } for tools that only read.