Evergreen Context Recall

Context Recall — "Refresh My Memory"

Works with Evergreen, a local-first personal CRM for macOS. Get it on the Mac App Store.

When to Use

• "What do I know about [name]?" before reaching out
• Someone emails you and you can't place them
• Before replying to a message from a contact you haven't talked to in a while
• When you need to brief someone else about a contact

How It Works

1. Look up the contact with get_contact for full profile
2. Pull the interaction history with get_contact_interactions, passing limit: 50 — that is the tool's maximum, not just a generous number, and the default of 10 drops the oldest interactions. The read takes no offset or cursor, so 50 is a hard ceiling: if exactly 50 come back, the history is truncated and the earliest interaction is not reachable through this API at all. Say so rather than presenting a partial timeline as complete
3. Check pending actions with list_actions, passing limit: 100 (its maximum)
4. Map their network with get_contact_network for relationships and introductions
5. Check introduction chains with get_introduction_chain
6. Compile into a narrative summary

Output Format

## [Name] — What You Know

**[Title] at [Organization]** | [Location]
**In your CRM since:** [date] | **Last interaction:** [date] ([days] ago)
**Tags:** [tags]

### How You Connected
[Narrative: who introduced you, where you met, initial context. If the interaction
read came back at its 50-entry maximum, the earliest interaction is missing and
cannot be fetched — qualify this section ("the record starts at [date]") instead of
asserting how you met]

### Relationship Timeline
- **[Date]** — [Interaction type]: [Summary]
- **[Date]** — [Interaction type]: [Summary]
- **[Date]** — [Interaction type]: [Summary]
[... back to the earliest interaction returned; if 50 came back, note here that the
history is truncated at the tool's maximum and the true first interaction is not
retrievable]

### What They're Working On
[Synthesized from notes and recent interactions — their current focus,
projects, interests]

### Your Shared Network
- **Introduced by:** [name and context]
- **You both know:** [mutual connections]
- **You introduced them to:** [names]

### Open Threads
- [Pending action 1]
- [Pending action 2]
- [Unanswered question or unresolved topic from last interaction]

### Key Details
[Anything notable from their notes — preferences, interests, important
context for future conversations]

Example

Request: "Refresh my memory on Marcus Webb"

Output:

## Marcus Webb — What You Know

**Founder & CEO at DataFlow** | Atlanta, GA
**In your CRM since:** Sep 2025 | **Last interaction:** Apr 1, 2026 (4 days ago)
**Tags:** founder, data, atlanta, startup

### How You Connected
David Kim introduced you at the Atlanta AI Dinner in September 2025.
Marcus was building DataFlow out of stealth at the time.

### Relationship Timeline
- **Apr 1** — Email: You sent the partnership proposal for data pipeline collaboration
- **Mar 20** — Meeting: Deep-dive on data architecture at his office
- **Feb 10** — Coffee: Caught up on DataFlow's progress, discussed hiring challenges
- **Dec 5** — Meeting: Holiday dinner, met his co-founder Alex Torres
- **Sep 15** — Meeting: First met at Atlanta AI Dinner

### What They're Working On
Building a real-time analytics platform for enterprise data pipelines.
Previously spent 5 years at Google on BigQuery. Recently closed a
$4M seed round. Hiring engineers aggressively.

### Your Shared Network
- **Introduced by:** David Kim
- **You both know:** Sarah Chen (Meridian Health), David Kim (DataTech)
- **His co-founder:** Alex Torres (met at Dec holiday dinner)

### Open Threads
- Partnership proposal sent Apr 1 — awaiting response
- Promised to intro him to your contact at Stripe

### Key Details
- Prefers morning meetings (noted from scheduling patterns)
- Big fan of Rust and real-time systems
- Has a dog named Byte (mentioned multiple times)

Checklist

Context Recall:
- [ ] Full contact profile retrieved
- [ ] `get_contact_interactions` called with `limit: 50`, its maximum (never the default 10)
- [ ] If 50 interactions came back, the timeline is flagged as truncated and "How You Connected" is qualified rather than asserted — this read has no pagination, so the earliest interaction cannot be fetched
- [ ] Network and relationships mapped
- [ ] Introduction chain traced
- [ ] Open actions and threads identified
- [ ] Narrative is readable and useful (not just raw data)

Learn More

• Evergreen — Local-First Personal CRM
• Vibe Coding a Personal CRM
• Evergreen Gets Serious
• Evergreen Gets Even Evergreener