On this page
Reference
ctx_feedback
Input contract and parameters for ctx_feedback.
MCP tool · Analysis tools
Call this tool through your MCP client. Its availability depends on your active configuration and profile.
This registry contract is separate from the SDK. For terminal commands, use the CLI reference.
Parameters
Generated from the registered input schema. “Required” reflects the schema; individual actions can impose additional requirements.
actionstringrecord (log event) | report (summary) | json (data) | reset (clear) | status (storage)
Accepted: recordreportjsonresetstatus
agent_idstringAgent ID (default: current agent)
intentstringIntent/task string
latency_msintegerLatency in ms (for record)
limitintegerMax recent events (default: 500)
llm_input_tokensintegerRequired for action=record
llm_output_tokensintegerRequired for action=record
modelstringModel identifier
notestringNote (no prompts/PII)
Exact JSON schema
Inspect the complete input schema
{
"properties": {
"action": {
"description": "record (log event) | report (summary) | json (data) | reset (clear) | status (storage)",
"enum": [
"record",
"report",
"json",
"reset",
"status"
],
"type": "string"
},
"agent_id": {
"description": "Agent ID (default: current agent)",
"type": "string"
},
"intent": {
"description": "Intent/task string",
"type": "string"
},
"latency_ms": {
"description": "Latency in ms (for record)",
"type": "integer"
},
"limit": {
"description": "Max recent events (default: 500)",
"type": "integer"
},
"llm_input_tokens": {
"description": "Required for action=record",
"type": "integer"
},
"llm_output_tokens": {
"description": "Required for action=record",
"type": "integer"
},
"model": {
"description": "Model identifier",
"type": "string"
},
"note": {
"description": "Note (no prompts/PII)",
"type": "string"
}
},
"required": [],
"type": "object"
}Runtime description
The following is the tool's own agent-facing description, reproduced from the reviewed registry. It can contain implementation terminology; it does not expand the public stability commitment.
Read the agent-facing description
Record and report LLM token/latency metrics — use to track efficiency and optimize context usage.
WORKFLOW: action=record during each LLM call, then action=report for readable summary.
Actions: record (log event), report (readable summary), json (machine-readable),
reset (clear data), status (storage info).
ANTIPATTERN: not for debugging code behavior — this tracks token/latency stats only.
record requires llm_input_tokens + llm_output_tokens.Sources & versions
- mcp_manifest.rsrust/src/core/mcp_manifest.rsReviewed checkout
- registry.rsrust/src/server/registry.rsReviewed checkout
- dynamic_tools.rsrust/src/server/dynamic_tools.rsReviewed checkout
- Core checkout
50fbff09c9- Installed runtime
- 3.10.2
- SDK release
- 1.1.0
Separate baselines for source, CLI/configuration and SDK contracts. Review does not certify every platform or integration.
Versions & compatibility