Every new session shouldn’t mean starting over.
KodaCode builds lasting project knowledge from what your agents discover: how to build and test your code, which fixes worked, why architectural decisions were made, and the conventions your team follows. Before each turn, it retrieves the memories relevant to the task, giving your agents the context they need without making them relearn the project.
Explore memories as a list, graph, or timeline. Correct mistakes, connect related knowledge, remove outdated information, and trace every change through its history.
Commit project memory alongside your source code so teammates can share discoveries, preserve decisions, and give every agent the benefit of lessons learned across the team.
The longer you use KodaCode, the more your project’s knowledge compounds.
Ask the agent "why do webhooks fail in staging?" and it recalls the problem, its cause and the fix, together.
Browse memories as a list, as a graph of what solves and causes what, or as a timeline of when each was learned.
Change a title or its content, retag it, or raise its importance so it comes up first.
Join two memories with the relationship that explains them: solves, causes, builds on, replaced by and more.
Every change to a memory is kept, so you can see what it said before and who changed it.
Export to Markdown or JSON, or import a memory you've exported before.
Memory stays on your computer unless you set it to be shared. A shared memory is a file that you commit with the project. Every team member's KodaCode reads that file and adds to it. When one person's agent finds a fix, every agent on the team can use it.