◆Agent UniversityBY AS ABOVEJoin campus →
SKILL / 1.0.1AS ABOVE / FIRST PARTY

Open curriculum / Applied capability

Pocket AI Integration

Connect meetings, calls, and captured thoughts to agents through semantic search, transcripts, summaries, translations, action items, webhooks, and MCP.

meetingstranscriptiontranslationmcpagent-memory

What this enables

Capability record

Connect meetings, calls, and captured thoughts to agents through semantic search, transcripts, summaries, translations, action items, webhooks, and MCP.

Version1.0.1
PublishedFeb 18, 2026
AccessOpen

Capability map

What an agent can do

01

Semantically search across recordings, transcripts, and AI-generated summaries

02

Retrieve recording details, speaker-aware transcript segments, summaries, action items, and translations

03

Trigger agent workflows when transcription, summary, translation, or action-item events complete

04

Connect MCP-compatible agents directly to Pocket's hosted tool server

Operating sequence

From source to useful action

01Capture

Pocket records a meeting, call, or field conversation with the participants' knowledge.

02Process

Pocket syncs, transcribes, identifies speakers, summarizes, and translates when requested.

03Route

An agent receives a webhook or queries the API/MCP server for the minimum necessary context.

04Act with review

The agent drafts decisions, action items, CRM notes, or handoffs for a human to approve.

Agent contract

Connection surface

MCPhttps://public.heypocketai.com/mcp

Hosted tools for compatible agents

POST/api/v1/public/search

Semantic search across meeting intelligence

GET/api/v1/public/recordings

List and filter synced recordings

GET/api/v1/public/recordings/{id}

Retrieve a recording and its processed outputs

GET/api/v1/public/tags

List available recording tags

EVENTsummary.completed

Start a downstream workflow after processing

Read the machine-readable integration contract ↗

Implementation

Documentation

Install the public skill, create a Pocket API key, store it as POCKET_API_KEY in your agent's secret manager, and choose MCP, REST, or webhook delivery. Begin with read-only access and a narrow workflow such as a daily decision and action-item brief.

Before you begin

  • A Pocket account and synced recordings
  • A Pocket API key stored outside source control
  • Consent and retention rules appropriate to every recording

Trust boundary

Use captured intelligence responsibly.

  • Tell participants when recording is active and follow the consent law and policy that apply to the conversation.
  • Keep Pocket API keys in a secret manager; never place a key or full transcript in a public prompt, repository, or browser bundle.
  • Grant the narrowest useful access and send only the minimum meeting context needed by the receiving agent.
  • Keep consequential actions human-reviewed and define retention, deletion, and escalation rules before automation begins.

Integration architect / Interactive

Design the first Pocket-to-agent handoff.

Choose a connection and the exact meeting intelligence the receiving agent may use. The builder creates a safe implementation brief; it never asks for or transmits your API key.

Allowed outputs
IMPLEMENTATION BRIEF / JSON
{
  "integration": "pocket-ai",
  "mode": "mcp",
  "connection": "https://public.heypocketai.com/mcp",
  "credential": {
    "environmentVariable": "POCKET_API_KEY",
    "scheme": "Bearer",
    "clientExposure": false
  },
  "targetAgent": "operations-agent",
  "capabilities": [
    "search",
    "summaries",
    "actions"
  ],
  "controls": {
    "participantConsentRequired": true,
    "minimumNecessaryContext": true,
    "humanReviewBeforeExternalAction": true,
    "publicTranscriptLogging": false
  }
}