pumaDB

pumaDB

pumaDB, by an independent developer, offers agents durable memory without infrastructure. Connect hosted MCP or REST, store JSON rows, and review persistent memory for agent workflows.

agent memorydurable storageMCPREST APIJSON database
Open site
image for pumaDBOpen site

What is pumaDB?

pumaDB is a hosted memory service for AI agents, built by an independent developer. It provides durable, persistent memory without requiring users to set up or manage a database. You connect via a hosted MCP (Model Context Protocol) endpoint or a server-side REST API, store JSON rows, and review what your agents remember across sessions. It is designed to solve the problem of agents forgetting context, preferences, or task state.

Application scenarios

  • Agent memory persistence

    Store user preferences, communication style, or formatting defaults so agents recall them across conversations.

  • Project conventions

    Keep repository facts, architecture notes, branch rules, and naming patterns that agents should not rediscover each session.

  • Research clippings

    Save sources, summaries, links, and follow-up questions from ongoing investigations.

  • Task state management

    Track open threads, blockers, handoff notes, and pending commands for long-running workflows.

  • Reusable skills

    Store Markdown-based operating instructions, project-specific workflows, and tool notes that agents load across sessions.

  • Typed safe memory

    Store resources, code snippets, commands, and config examples as inert records for review before use.

Core Features

  • Hosted MCP endpoint

    Connect from Codex, ChatGPT, Claude, OpenClaw, or any client that supports Streamable HTTP MCP, with OAuth support for authentication.

  • Server-side REST API

    For small server-side apps, keep a named `puma_live_*` key in a backend or serverless environment and call the `/v1` API directly.

  • Lightweight memory schema

    Use pumaDB for almost any small, durable JSON record—facts, state, notes, settings, scratchpads, or handoff context.

  • Consolidated remember tool

    The MCP tool stores common memory types with inert safety metadata, keeping memory organized.

  • Safety rails

    Scoped table limits, per-key rate limits, filtered cleanup, natural edits, and automatic version history for updates and deletes.

  • Account limits

    20 tables with 1,000 rows per table, keeping memory deliberately small and explicit.

  • Quick setup

    No database project required—pick hosted MCP for agents or REST for server-side apps, then start writing memory in seconds.

Target users

pumaDB is for developers and AI engineers building agent-based applications. It suits anyone using Codex, ChatGPT, Claude, or other MCP-compatible clients who need durable memory without managing infrastructure. It also works for small server-side apps or serverless environments that require persistent JSON storage for agent workflows.

How to use pumaDB?

  1. Choose your connection method: hosted MCP for agents (via Streamable HTTP endpoint at https://api.pumadb.ai/mcp) or server-side REST API (using a puma_live_* key and calling /v1).
  2. For MCP, add pumaDB as a hosted server in your client (e.g., Codex, ChatGPT, Claude) using the provided URL and OAuth support.
  3. For REST, keep the API key in a backend or serverless environment—do not expose it in client-side code.
  4. Start writing memory using the pumadb.remember() tool and query it later with pumadb.query().

Effect review

pumaDB delivers exactly what it promises: durable memory for agents without database overhead. The dual MCP and REST support makes it flexible for both agentic and server-side workflows, while the safety rails and version history add practical safeguards. The 20-table, 1,000-row limit keeps memory scoped and manageable for most use cases. For developers tired of agents forgetting context, this is a clean, no-fuss solution that works out of the box.

Frequently Asked Questions

What is pumaDB?
pumaDB is a durable memory service for AI agents, allowing JSON data storage and retrieval via MCP or REST APIs, without requiring infrastructure management.
How do I connect to pumaDB?
You can connect using hosted MCP (Model Context Protocol) or REST APIs, both provided by pumaDB for seamless integration.
What type of data can I store?
pumaDB stores JSON rows, enabling flexible and structured data storage for agent workflows.
Is pumaDB suitable for production use?
Yes, it is designed for production agent workflows with persistent memory and reliable data storage.
Do I need to manage my own infrastructure?
No, pumaDB is fully hosted, so you can focus on building agents without worrying about servers or databases.
Who develops pumaDB?
pumaDB is developed by an independent developer, ensuring focused and dedicated support for the tool.

pumaDB - AI Tool Detail

pumaDB, by an independent developer, offers agents durable memory without infrastructure. Connect hosted MCP or REST, store JSON rows, and review persistent memory for agent workflows.

Category：Agents

Visit Link：https://pumadb.ai/

Tags：agent memory、durable storage、MCP、REST API、JSON database