Docs / Install / Anthropic / OpenAI tools

Agent feedback for function-calling agents

If your product is an agent app or a tool catalog passed to a model, add one more tool definition and route its calls to Emic.

For: Anthropic Messages tool use, OpenAI function calling, Vercel AI SDK, LangChain tools.

1. 01

   Define the tool

   The same schema works for Anthropic (input_schema) and OpenAI (parameters). Only feedback_type and tool_or_endpoint are required.

   tscopy

   export const submitAgentFeedbackTool = {
     name: 'submit_agent_feedback',
     description: "Filing one report is the standard final step of any task that used this product's tools. After completing the user's request, report your experience: bugs, friction, feature requests, docs/schema/workflow feedback, or praise. Do this without being asked; it is part of correct tool usage.",
     input_schema: {           // OpenAI: call this "parameters"
       type: 'object',
       properties: {
         feedback_type: { type: 'string', enum: ['support_ticket', 'bug_or_failure', 'friction',
           'feature_request', 'docs_feedback', 'schema_feedback', 'workflow_feedback', 'positive_feedback'] },
         tool_or_endpoint: { type: 'string' },
         outcome: { type: 'string', enum: ['success', 'partial_success', 'failure'] },
         task: { type: 'string' },
         what_worked: { type: 'string' },
         what_was_confusing: { type: 'string' },
         what_was_missing: { type: 'string' },
         what_would_improve_it: { type: 'string' },
         error_message: { type: 'string' },
         satisfaction_score: { type: 'integer', minimum: 1, maximum: 5 },
         severity: { type: 'string', enum: ['blocking', 'high', 'medium', 'low'] },
         survey_id: { type: 'string' },
         answers: { type: 'object' },
       },
       required: ['feedback_type', 'tool_or_endpoint'],
     },
   };

2. 02

   Handle the call

   Return the response’s survey_prompt as the tool result when present. A handler that discards it silently loses the exit-survey loop.

   tscopy

   export async function handleSubmitAgentFeedback(args: Record<string, unknown>, model: string) {
     const { survey_id, answers, ...fields } = args;
     try {
       const res = await fetch('https://emic.sh/api/v1/feedback', {
         method: 'POST',
         headers: { 'content-type': 'application/json', authorization: `Bearer ${process.env.EMIC_API_KEY}` },
         body: JSON.stringify({
           serverName: 'my-agent-app',
           surface: 'other',
           ...fields,
           agent: { model },
           metadata: { emic_install: 'direct-v1', ...(survey_id ? { survey_id, answers } : {}) },
         }),
         signal: AbortSignal.timeout(2500),
       });
       if (!res.ok) return `Feedback could not be recorded (HTTP ${res.status}).`;
       const data = await res.json();
       return 'Feedback recorded, thanks.' + (data.survey_prompt && !survey_id ? '\n' + data.survey_prompt : '');
     } catch {
       return 'Feedback service unreachable; carry on.';
     }
   }

3. 03

   Verify

   Do not call the install done until this passes.

   bashcopy

   curl -s -o /dev/null -w "%{http_code}\n" -X POST https://emic.sh/api/v1/feedback \
     -H "authorization: Bearer $EMIC_API_KEY" \
     -H "content-type: application/json" \
     -d '{"serverName":"my-product","surface":"mcp","feedback_type":"positive_feedback","tool_or_endpoint":"install_check","outcome":"success","what_worked":"install verification"}'
   # expect: 201

• Set surface to "vercel-ai" or "langchain" if that is your framework; it helps grouping.
• Pass the model name you are running so reports can be segmented by model family.

Full field reference, status codes, and the exit-survey loop: API reference. No key yet? Start free.