Stash Overview
Stash does two things: memory for the coding agents your team runs, and memory for the agents inside your product.
Internal agents: self-improving skills
The Stash plugin records the sessions your coding agents already produce — Claude Code, Cursor, Codex, and the rest — into one shared, searchable store. A nightly curator refines them into a wiki of pages and reusable skills, and every agent reads that memory back through the CLI or MCP on its next run. One engineer's debugging session becomes something every teammate's agent already knows.
External agents: the Developer Platform
The same memory loop for the agents inside your product. Your agents push their conversations to Stash per end user; the curator writes each user a private wiki plus one shared, anonymized wiki that every user's agent reads. Your product gets smarter with every conversation, without you building a memory system.
Example: Don't Duplicate Work
Henry asks his coding agent to investigate a memory leak. His teammate Sam already spent hours debugging the same issue the night before. Without Stash, the agent starts from scratch. With Stash, it picks up where Sam left off.
Example: Managing Upwards
After a long day of working with coding agents, I ask “what did I get done today?”
FAQ
Do I have to upload my transcripts?
No. Session recording is on by default during setup, but it's entirely yours to control: decline it in the wizard, pause globally with stash stop, pick which agents record, or exclude folders in stash settings. Your coding agent keeps read access to your Stash either way.