pumaDB
Hosted memory layer for AI agents, with JSON rows and version history
Visit website →About
pumaDB is a small hosted memory layer that gives AI agents a persistent place to store user preferences, project conventions, research notes, and task state — without forcing the developer to run a database. It exposes both an MCP server and a REST API, so agents reach the same store regardless of how they're orchestrated.
Each row is a JSON blob with a key, automatic version history, and safety hooks for soft-delete and rollback. The free tier covers 20 tables and 1,000 rows per table, which is enough to ship a working agent and decide later if scale warrants migration.
The product targets teams who keep redesigning a memory tier on every agent project. By pinning durability, schema flexibility, and history to one hosted endpoint, pumaDB lets the agent focus on reasoning instead of state management.
AI-assisted draft, human-reviewed before publishing — see how we choose & review tools.
Why we picked it
pumaDB earns its spot for exposing both MCP and REST to the same store — reaching the same memory regardless of how an agent is orchestrated is worth naming for a team running a mix of agent frameworks that all need to see the same persisted state.
Related tools
- 1 Cekura Test voice agents before launch, monitor after, improve automatically Freemium
- 2 Databench Open-source agentic workspace for data teams — collaborate with AI agents Freemium
- 3 Statable Analytics Privacy-friendly Google Analytics alternative with MCP support for agents Trial
- 4 Anomalo Analyst AI agents that monitor your data and surface what matters before you ask Free
- 5 TrackMCP Analytics for MCP servers — one line of code. Freemium
- 6 Olostep Web data infrastructure for AI agents Freemium