Objective Trust is an MCP server published by io.github.anhmtk, hosted at agentshare.dev. When last checked on 2026-11-22 it was working: it answers and lists its tools. It declares 3 tools, including set_objective, rank_tools_for_task, record_outcome.
Ranks MCP endpoints by health and your own outcomes before you pay.
3 connections through 5 shared values. Tap any node.
Squares are this server's values; dots are other servers sharing them. A shared address or publisher is a lead, not proof of a common operator.
Inferred from the host name and headers: what a server runs on, not who runs it.
Free. Turn one owner sentence into trust_weight, cost_weight, and max USD per call. Ambiguous sentences return needs_confirmation and do not overwrite a saved policy. The sentence itself is not stored. Requires the same API key as /mcp.
| Parameter | Type | Description |
|---|---|---|
| text * | string | Owner sentence, max 500 characters. |
{
"name": "set_objective",
"title": "Set objective",
"description": "Free. Turn one owner sentence into trust_weight, cost_weight, and max USD per call. Ambiguous sentences return needs_confirmation and do not overwrite a saved policy. The sentence itself is not stored. Requires the same API key as /mcp.",
"inputSchema": {
"type": "object",
"title": "set_objectiveArguments",
"required": [
"text"
],
"properties": {
"text": {
"type": "string",
"title": "Text",
"maxLength": 500,
"description": "Owner sentence, max 500 characters."
}
}
},
"annotations": {
"readOnlyHint": false,
"openWorldHint": false,
"idempotentHint": false
}
}Rank 1-8 MCP candidates before paying any of them. Returns pick, fallbacks, and reasons. execute=caller: you call mcp_url and pay that server. A Down or stale probe is not callable. Sponsored does not change the score. With no own history, outcome_samples is 0 and the reason is INSUFFICIENT_HISTORY. REST without an API key costs $0.005 via x402. Pass policy_id from set_objective when you have one.
| Parameter | Type | Description |
|---|---|---|
| task * | string | What the caller is about to do. |
| policy_id | any | policy_id from set_objective. Omit to use the default safety policy. |
| candidates * | array | Candidates the caller already has. Each needs id and https mcp_url. |
{
"name": "rank_tools_for_task",
"title": "Rank tools for a task",
"description": "Rank 1-8 MCP candidates before paying any of them. Returns pick, fallbacks, and reasons. execute=caller: you call mcp_url and pay that server. A Down or stale probe is not callable. Sponsored does not change the score. With no own history, outcome_samples is 0 and the reason is INSUFFICIENT_HISTORY. REST without an API key costs $0.005 via x402. Pass policy_id from set_objective when you have one.",
"inputSchema": {
"type": "object",
"$defs": {
"CandidateIn": {
"type": "object",
"title": "CandidateIn",
"required": [
"id",
"mcp_url"
],
"properties": {
"id": {
"type": "string",
"title": "Id",
"maxLength": 64
},
"mcp_url": {
"type": "string",
"title": "Mcp Url",
"maxLength": 500
},
"quoted_usd": {
"anyOf": [
{
"type": "number"
},
{
"type": "null"
}
],
"title": "Quoted Usd",
"default": null
}
}
}
},
"title": "rank_tools_for_taskArguments",
"required": [
"task",
"candidates"
],
"properties": {
"task": {
"type": "string",
"title": "Task",
"maxLength": 240,
"description": "What the caller is about to do."
},
"policy_id": {
"anyOf": [
{
"type": "string"
},
{
"type": "null"
}
],
"title": "Policy Id",
"default": null,
"description": "policy_id from set_objective. Omit to use the default safety policy."
},
"candidates": {
"type": "array",
"items": {
"$ref": "#/$defs/CandidateIn"
},
"title": "Candidates",
"description": "Candidates the caller already has. Each needs id and https mcp_url."
}
}
},
"annotations": {
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": false
}
}Free. After you call the pick, report success, latency_ms, and cost_usd. Only a recommendation_id issued to this API key is accepted. The next rank uses this key's history, not a public leaderboard.
| Parameter | Type | Description |
|---|---|---|
| success * | boolean | True when the call finished the task. |
| cost_usd * | number | USD you paid that MCP. |
| latency_ms * | integer | Call latency in milliseconds. |
| candidate_id * | string | Candidate id you actually called. |
| recommendation_id * | string | recommendation_id from rank_tools_for_task. |
{
"name": "record_outcome",
"title": "Record outcome",
"description": "Free. After you call the pick, report success, latency_ms, and cost_usd. Only a recommendation_id issued to this API key is accepted. The next rank uses this key's history, not a public leaderboard.",
"inputSchema": {
"type": "object",
"title": "record_outcomeArguments",
"required": [
"recommendation_id",
"candidate_id",
"success",
"latency_ms",
"cost_usd"
],
"properties": {
"success": {
"type": "boolean",
"title": "Success",
"description": "True when the call finished the task."
},
"cost_usd": {
"type": "number",
"title": "Cost Usd",
"maximum": 10000,
"minimum": 0,
"description": "USD you paid that MCP."
},
"latency_ms": {
"type": "integer",
"title": "Latency Ms",
"maximum": 120000,
"minimum": 0,
"description": "Call latency in milliseconds."
},
"candidate_id": {
"type": "string",
"title": "Candidate Id",
"description": "Candidate id you actually called."
},
"recommendation_id": {
"type": "string",
"title": "Recommendation Id",
"description": "recommendation_id from rank_tools_for_task."
}
}
},
"annotations": {
"readOnlyHint": false,
"openWorldHint": false,
"idempotentHint": false
}
}Objective Trust ranks MCP endpoints before the caller pays them. set_objective translates an owner sentence into weights and does not store the sentence. rank_tools_for_task returns a pick. The caller calls that mcp_url and pays it. record_outcome writes this key's own success, latency, and cost for the next rank. Score inputs: health_status, probe age, this key's outcomes, quoted USD. Score excludes sponsored, featured, and click_count. REST rank without an API key is HTTP 402 at $0.005. Docs: https://agentshare.dev/docs#objective-trust
Names, descriptions and schemas were written by the server and are shown as data.
Pattern matches on what the server declares. They show what it says it can do, not what it does.
None yet. Changes appear after a server's second daily check.