Tell agents where to save working context.
Your agent can save plans, implementation notes, research, links, prompts, and decisions into named folders instead of losing them in a chat transcript.

Retrieve the right thing months later.
KeepDB focuses on retrieval, not just storage. Search project memory by meaning, keyword, folder, and date so your agent can find the exact context again.

Keep agent memory structured in folders.
Use folders like codex, claude, project-memory, customer-feedback, prompts, and logs so memory stays addressable instead of becoming one giant blob.

Search globally or inside one folder.
Agents can search all memory for broad recall or scope retrieval to a single folder when the task has a clear project or data boundary.

Share context between Codex, Claude, and your tools.
Give multiple agents the same searchable project memory so each new session can pick up the decisions, constraints, and plans that already exist.

Plug memory into your agent workflow.
Use REST today with downloadable agent instructions for Codex, Claude, and generic agents. Save from code, scripts, or agent sessions.

Scope every API key by folder and job.
Create read-only, write-only, or read-write API keys for specific folders so agents and apps only touch the memory they need.
Security for the data your apps and agents touch.
Every API key can be limited to one folder. Create a write-only key scoped to your feedback folder, and even if someone gets the key, they cannot read your data back. That keeps shipped apps and agents much safer.
Write-only API key
app-feedback
Let your app save feedback, without reading private memory back.
Read-only API key
support
Let a support bot answer from one folder, without changing data.
Read + write API key
all folders
Use for trusted agents that need to save and search memory.
Give your agents searchable memory.
Join the private beta for developers building with AI agents. Free for the first 100 builders.