# AI search MCP and API: connect evidence, execution and outcomes. - working worksheet

Reviewed alongside the Eli guide: https://hireeli.io/resources/ai-search-mcp-api

## Decision to make

Evaluate an MCP server or API for AI search monitoring and execution

Owner: ____________________
Review date: ____________________
Canonical page or destination: ____________________

## Buyer questions

- [ ] When should a team use MCP instead of REST?
- [ ] Should an MCP client be able to publish immediately?
- [ ] Does Eli publish an MCP and API reference?

## Evidence-led work plan

### 1. Read evidence before choosing an action

The client should be able to inspect the workspace, measured question, provider response, source evidence and current priority without receiving publishing access.

Evidence or artifact: ____________________
Status: [ ] not started  [ ] working  [ ] verified  [ ] blocked

### 2. Separate preparation, approval and execution

A useful integration can prepare an exact page change or buyer-journey action without publishing it.

Evidence or artifact: ____________________
Status: [ ] not started  [ ] working  [ ] verified  [ ] blocked

### 3. Connect to the business outcome without merging evidence

The same API can expose AI answers, site publications, AI-referred activity, qualified actions and CRM-confirmed outcomes, but these records should keep their own provenance.

Evidence or artifact: ____________________
Status: [ ] not started  [ ] working  [ ] verified  [ ] blocked

### 4. Evaluate an MCP tool as an auditable interface

Inspect the tool schema before connecting an agent.

Evidence or artifact: ____________________
Status: [ ] not started  [ ] working  [ ] verified  [ ] blocked

## Source ledger

- [ ] Model Context Protocol specification: https://modelcontextprotocol.io/specification/latest
- [ ] OAuth 2.0 authorization framework: https://www.rfc-editor.org/rfc/rfc6749
- [ ] Eli MCP and API documentation: https://hireeli.io/docs
- [ ] Eli: AI Visibility Index methodology: https://hireeli.io/ai-visibility-index/methodology

## Measurement boundary

Baseline date and exact context: ____________________
7-day observation: ____________________
14-day observation: ____________________
30-day observation: ____________________
Known limitations: ____________________

Keep technical eligibility, citations, recommendations, visits, leads and revenue as separate evidence. This worksheet does not guarantee an external ranking or commercial result.