Skip to main content
AI/MLjeremylongshore

fathom-reference-architecture

'Reference architecture for Fathom meeting intelligence integrations.

Stars
2,267
Source
jeremylongshore/claude-code-plugins-plus-skills
Updated
2026-05-31
Slug
jeremylongshore--claude-code-plugins-plus-skills--fathom-reference-architecture
View on GitHubRaw SKILL.md

// install — copy + paste into any project

mkdir -p .claude/skills && curl -fsSL https://raw.githubusercontent.com/jeremylongshore/claude-code-plugins-plus-skills/HEAD/plugins/saas-packs/fathom-pack/skills/fathom-reference-architecture/SKILL.md -o .claude/skills/fathom-reference-architecture.md

Drops the SKILL.md into .claude/skills/fathom-reference-architecture.md. Works with Claude Code, Cursor, and any agent that loads SKILL.md files from .claude/skills/.

Fathom Reference Architecture

Overview

Define meeting ingestion, consent, Fathom processing, CRM/follow-up integration, redacted observability, retention, and incident boundaries as an owned system.

Prerequisites

  • A documented data/consent model, environment boundaries, integration owners, and approved data-retention policy.

Instructions

  1. Map meeting sources, identities, processing, integrations, access, and audit paths to owners.
  2. Separate development/staging/production credentials and use role-limited access at every boundary.
  3. Build idempotent integrations and fallback/rollback decisions before production automation.

Output

  • An architecture record with trust boundaries, ownership, consent/data controls, and reversible integration points.

Error Handling

Condition Safe response
CRM or follow-up integration fails Pause the affected automation and use the documented reconciliation path.
Access boundary is unclear Use the restrictive setting and escalate to the data/tenant owner.
Meeting content is exposed Restrict access and follow the incident procedure.

Examples

Model a development flow from a synthetic meeting to a scoped Fathom integration and test CRM record, retaining only opaque correlation/audit metadata. Promote the reviewed version through staging before a production canary; do not make summary output an unreviewed system of record.

Architecture

┌──────────────┐     ┌─────────────────┐     ┌──────────────────┐
│  Fathom AI   │────▶│  Webhook        │────▶│  Meeting DB      │
│  (Recordings)│     │  Handler        │     │  (PostgreSQL)    │
└──────────────┘     └─────────────────┘     └────────┬─────────┘
                                                       │
                     ┌─────────────────┐     ┌────────▼─────────┐
                     │  Action Item    │     │  CRM Sync        │
                     │  Extractor      │────▶│  (Salesforce/    │
                     └─────────────────┘     │   HubSpot)       │
                            │                └──────────────────┘
                     ┌──────▼──────────┐
                     │  Follow-up      │
                     │  Email Sender   │
                     └─────────────────┘

Project Structure

fathom-platform/
├── src/
│   ├── fathom_client.py
│   ├── webhook_handler.py
│   ├── transcript_processor.py
│   ├── action_extractor.py
│   ├── crm_sync.py
│   └── email_sender.py
├── sql/
│   └── schema.sql
├── tests/
│   ├── fixtures/
│   └── test_processor.py
└── deploy/
    ├── cloud-function/
    └── docker-compose.yaml

Key Design Decisions

Decision Choice Rationale
Data delivery Webhooks Real-time, no polling
Storage PostgreSQL Structured meeting data
Processing Cloud Function Serverless, scales with meeting volume
CRM sync Async queue Handles CRM rate limits

Resources

Next Steps

This completes the Fathom skill pack. Start with fathom-install-auth.