As a Network Admin, you can add memories that help Coworker understand your organization and how it uses specific connectors.
What are network and connector memories?
Memories are persistent pieces of contextual information that Coworker automatically references when running Deep Work tasks. Unlike Skills, which are reusable instruction modules you attach to Agents, Memories are contextual facts that help Coworker understand how your organization works.
There are four types of memories:
Company Context — Broad organizational knowledge that applies to all users on your network
Network Memory — Specific to your Coworker network configuration and defaults
Connector Memory — Describes how your organization uses specific data connectors (e.g., Jira, HubSpot, Salesforce)
Personal Memory — Individual user preferences and context
This article focuses on Company Context, Network Memory, and Connector Memory — all of which can be created and managed by Network Admins.
Prerequisites
You must be a Network Admin for your Coworker network
The relevant data connectors must already be connected at the network level
Why connector memories matter
Every organization uses tools differently. Your Jira instance has custom fields, your HubSpot pipeline has specific stages, and your Slack channels follow unique naming conventions. Without connector memories, Coworker has to guess or ask for clarification every time it queries these tools.
Connector memories solve this by teaching Coworker:
Which Jira projects house which types of work
What custom fields mean in your HubSpot or Salesforce instance
How your Slack channels are organized by department or function
Which Confluence spaces store what kinds of documentation
Nuances in your data sources that aren't visible through standard API access
How to create network and connector memories
Step 1: Navigate to the Memory settings
Open Coworker and click the gear icon (Settings) in the bottom-left corner
Select Memories from the settings menu
You will see tabs for Company Context, Network Memory, and Connector Memory
Step 2: Choose your memory type
Click the tab for the type of memory you want to create:
Company Context — For general organizational knowledge (e.g., "We use HubSpot as our source of truth for customer data, not Salesforce")
Network Memory — For network-specific defaults and configurations
Connector Memory — For tool-specific context and customizations
Step 3: Add your memory
Click Add Memory or the plus icon
Enter a clear, descriptive title for the memory
Write the memory content in plain language. Be specific and factual
Click Save
Step 4: Verify injection (optional)
Once saved, your memory will be automatically injected into all chat prompts across your network. You can verify this by starting a new chat and checking that Coworker references the context correctly.
Best practices for writing memories
Be specific — "We store customer data in HubSpot" is better than "We use CRM tools"
Include IDs and codes — When relevant, include project keys, pipeline IDs, or channel IDs
Explain custom fields — Document what custom fields mean and how they're used
Note source-of-truth hierarchies — If HubSpot conflicts with a spreadsheet, state which wins
Keep it concise — Aim for 1-3 sentences per memory. Longer content can be split into multiple memories
Update regularly — As your tools and processes evolve, revisit and refresh your memories
Example connector memories
Here are examples of effective connector memories for common tools:
Jira
"Our primary project key is AZ. Tickets with status 'DONE (READY FOR CUSTOMERS)' are considered closed."
"Custom field 'Customer Priority' uses values: P0 (critical), P1 (high), P2 (medium), P3 (low)."
HubSpot
"Our main deal pipeline ID is 987654321. Deal stages in order: Discovery, Solution Evaluation, POC Planning, Security Review, Technical Implementation, Training, POC Execution, Validation, Final Agreement, Closed Won, Closed Lost."
"We use HubSpot as the source of truth for customer status. If a spreadsheet says something different, trust HubSpot."
Slack
"#engineering is used for general engineering discussions. #bugs-general-discussion is for bug triage."
"Release notes are posted in #tech-announcements (channel ID: C1234567890)."
Confluence
"The 'Product' space contains roadmap and PRD documentation. The 'Engineering' space contains runbooks and technical specs."
Notes & Tips
Memories are network-wide — Once created, they apply to all users on your network during Deep Work sessions
They complement, not replace, Skills — Memories provide context; Skills provide instructions. Both work together
You can edit or delete memories anytime — Return to the Memory settings and click the memory you want to modify
Start with 3-5 high-impact memories — Focus on the tools and workflows that cause the most confusion or inconsistency