The 3rd-most used agent at Runlayer creates a living graph of our entire company. We couldn't operate without it.
How it works (bookmark this):
- Pulls from meeting transcripts, publicly accessible emails, public Slack channels, Linear tickets, and GitHub commits on a daily basis
- Reads this activity and spawns sub-agents to record it all
- Generates a knowledge graph of the company's decisions, action items, blockers, and key updates
- Inserts this into a living database that's constantly updating
The core idea is to record Runlayer's core systems and activity across the graph.
Then, synthesize that information to help with higher-level work.
A few sub-agents we run using this context graph:
- Product overview agent: updates a product overview page based on shipped work so it reflects what Runlayer offers, day-of, for GTM
- Messaging agent: continuously updates a Notion page with core messaging it hears across sales calls that is resonating
- Competitive battlecard agent: detects competitor mentions from sales calls, Slack, and emails, adds missing competitors, and updates battlecards with reasons they position themselves as better
Now, we have a constantly-updating database which employees (and agents) can reference.
We can learn from this memory layer to make decisions, prioritize features, and find out what repetitive tasks should become an agent.
Runlayer gives our employees a golden path to build agents like these, and this is the direct result.
Follow if you're interested in seeing more agents - I'll be publishing stuff like this every week. I'll probably post the top 2 next week.
