Personal CRM

Personal CRM

Summary

A personal CRM is a private system for remembering relationships: who someone is, how you know them, when you last spoke, what context matters, and what follow-up should happen next.

In this page, personal CRM means more than a contact list and less than a general AI context platform. The scope is relationship memory: contacts, communication history, cadence, reminders, personal notes, local data ownership, and carefully reviewed inference over private relationship context.

The Spencer Graham fireside and the PersonalCRM repository provide a concrete source anchor for this topic. They should not be treated as the whole category. Monica, an open-source personal CRM, shows that the category exists beyond this single project. Local-first software provides the adjacent frame for why data ownership and offline-capable operation matter.

Definition

A personal CRM is software for maintaining a personal or professional relationship graph. It usually combines contact records, notes, reminders, interaction history, and prompts that help a person stay in touch.

The durable wiki scope is:

who the person is

how they can be contacted

when interaction happened

what follow-up is open

what relationship context should be remembered

what data should remain under the user's control

The page should not collapse into a broader Context Systems page. Context ingestion, retrieval, AI memory, and agent coordination are related topics, but they need their own source map.

Why Personal CRM Matters

Most relationship context lives in scattered places: email, calendars, notes, chats, call history, meeting notes, and memory. A personal CRM tries to turn that scattered context into a usable relationship record.

For a builder or operator, the important jobs are practical:

remember why a relationship matters

avoid dropping follow-ups

prepare before a conversation

notice when cadence has drifted

preserve context without giving every platform permanent control over it

The privacy stakes are higher than in a normal task manager. A personal CRM may touch messages, meetings, personal notes, and inferred facts about other people. That makes consent, provenance, review, and local control part of the topic, not optional implementation details.

Relationship Memory

The core object in a personal CRM is not only a contact. It is the relationship around that contact.

A useful relationship-memory system may include:

contact methods such as email, phone, and social handles

notes and tags

organizations and roles

last-contacted timestamps

follow-up tasks

reminders or cadence rules

interaction timelines

source-linked facts and inferred context

PersonalCRM's architecture and codebase support this frame. Its repository includes contact and contact-method code, cadence/follow-up logic, contact tasks, external contact matching, Google integrations, Telegram paths, Apple Messages/Mac daemon paths, iCloud contact paths, and shared communication-message storage.

The page should be careful with graph language. PersonalCRM's README and specs point toward graph view, graph data modeling, and richer relationship inference, but those should be treated as roadmap or design direction unless later implementation evidence verifies them as shipped.

Data Ownership And Local-first Operation

PersonalCRM describes itself as a single-user, local-first CRM. Its architecture guide frames the system as desktop-first, local-first, and cloud-optional, optimized for personal use on constrained hardware.

That matters because personal CRM data can be sensitive. Relationship memory is not just business metadata. It can include communication history, personal context, meeting summaries, task commitments, and inferred facts. A local-first design gives the user more control over where the source of truth lives and which services are allowed to process it.

The Ink & Switch local-first essay is useful related reading here: it names user ownership, offline-capable operation, privacy, preservation, and control as core local-first concerns. The Personal CRM page should link local-first software as a related topic rather than trying to re-explain the whole field.

Communication Context And Cadence

Personal CRM becomes more useful when communication context updates relationship state. The source packet for PersonalCRM points to code and specs around Gmail, Google Calendar, Google Contacts, Google Chat, Telegram, Apple Messages, iCloud contacts, and shared communication-message storage.

A careful claim is that PersonalCRM has current code paths and design documents around these source areas. A less safe claim would be that every source is complete in every direction.

The important distinction is between contact methods and message ingestion. A contact method, such as a phone number or handle, does not prove that the system ingests messages from that platform.

One concrete caveat is open issue #489 in the PersonalCRM repository: outbound iMessages/SMS are not yet emitted by the live Messages source. That means the page should avoid saying that Apple Messages support is complete bidirectional relationship memory.

AI-assisted Extraction And Review

PersonalCRM's LLM extraction spec describes a direction for turning raw relationship content into useful relationship state. The planned jobs include task extraction, dynamic cadence, drift-with-reason, living relationship profiles, and briefings before interaction.

This belongs in the Personal CRM page as future/current direction, not as a shipped claim.

The key principle is review. The LLM extraction spec emphasizes trust, provenance, and human review. It also scopes out autonomous outward action. The system should surface tasks, nudges, corrections, and briefings; it should not be described as sending messages or acting on behalf of the user.

This is the point where Personal CRM touches the future Context Systems page. Context Systems should cover the broader infrastructure question: ingestion, retrieval, memory, provenance, verification, and coordination context for agents. Personal CRM is one private relationship-memory use case inside that wider problem.

PersonalCRM As An Implementation Example

PersonalCRM is a public repository by Spencer Graham. Its README describes it as a single-user, local-first CRM with AI-powered insights. As of the 2026-06-11 source scan, GitHub reports Go as the largest language footprint, with Swift and TypeScript also present.

Useful implementation facts for the wiki page:

core CRM work is in progress

contacts and contact methods are first-class implementation areas

cadence and follow-up logic are present in the backend

Google, Telegram, Apple Messages, and iCloud contact source areas appear in code or specs

communication-message storage exists in the backend

AI extraction, graph view, RAG/chat UI, and advanced inference should be framed as roadmap/spec direction unless separately verified

The page should use PersonalCRM as a concrete example, not as the definition of the category.

Boundaries And Caveats

Do not claim:

Personal CRM and Context Systems is an established single term

WhatsApp, Signal, or Discord message ingestion is implemented

graph view, RAG/chat UI, MCP briefing, or LLM extraction are shipped

outbound iMessages/SMS are currently emitted by the live Messages source

the system takes autonomous outward action on behalf of the user

Do say:

Personal CRM is a useful category for private relationship memory

PersonalCRM is one current implementation example

local-first architecture is relevant because relationship context can be sensitive

AI-assisted extraction needs provenance, review, and clear boundaries

Related Topics

Local-first Software

Personal Knowledge Management

Private Communication Archives

Relationship Graphs

Context Systems

Agentic Coding Workflows

UX QA Agents

Atomized Developer

Open Source Security in the Agentic Coding Era

Further Reading

PersonalCRM repository: https://github.com/spengrah/PersonalCRM

Monica: https://www.monicahq.com/

Monica GitHub repository: https://github.com/monicahq/monica

Local-first Software, Ink & Switch: https://www.inkandswitch.com/essay/local-first/

June Cohort Fireside Chats (Spencer Graham): https://portal.raidguild.org/sessions/49

Agentic Coding Works Better as a Doer/Reviewer Pipeline: https://portal.raidguild.org/posts/agentic-coding-doer-reviewer-pipeline

AI and Open Source Security in the Agentic Coding Era: https://portal.raidguild.org/wiki/ai-and-open-source-security-agentic-coding-era

Open Questions

Which PersonalCRM source integrations are complete end-to-end?

How should a personal CRM handle consent when communication history includes other people?

Which inferred relationship facts should require human review before being saved?

Should relationship graph material become a future child page?

Where should the boundary sit between Personal CRM and Context Systems?