Reclaw is a purpose-built database for AI coding agents like OpenAI Codex, Claude Code, and similar tools. Give your agents a real place to store, organize, and share the data they create.
Install a skill. Get durable memory for your agents.
AI agents are great at generating code, files, and workflows. They are not great at managing durable, structured application data across sessions, tasks, and teams. Reclaw fixes that.
Powered by a SQLite-style database that gives agents fast, reliable access to records.
Local-first speed with cloud-grade resilience for durability and recovery.
Collaborate on agent-generated data with a Dropbox-like experience.
Agents discover and understand structure simply by browsing the local filesystem.
Install a skill. Your agent starts persisting structured data right away.
Designed for autonomous agents working directly from the local filesystem—not retrofitted from human tools.
Context windows are not databases. Chat logs are not systems of record. Scratch files are not collaboration tools.
When agents create useful data—task logs, extracted entities, inventories, customer notes, research artifacts, generated metadata, workflow outputs—you need that data to be:
Reclaw delivers all five.
Agents like Codex or Claude Code navigate a local project directory, inspect files, infer structure, and take action. Reclaw embraces that behavior. Each table is represented by a clearly named file, making the schema easy for an agent to discover just by exploring the local filesystem.
Five steps from install to durable, team-shared agent memory.
Your agent gains access to a local-first structured data layer. No heavy setup, no complex provisioning.
Reclaw stores agent-created data in a local database built on a SQLite-style model—fast reads, minimal overhead.
Records are automatically backed up so your data stays durable, restorable, and safe.
Collaborate around agent-generated data with a familiar, Dropbox-style sharing model.
One file per table with human-readable names. Agents understand the schema by browsing the filesystem.
A persistent layer for structured data that survives beyond a single run, prompt, or session.
Agents work with data locally, where it is fastest and easiest to access.
Automatic S3 backups protect your data and make recovery straightforward.
Share records across your team so people and agents stay aligned on the same source of truth.
Agents inspect the local directory and understand the schema with minimal friction.
Install a skill and start using Reclaw without a complicated deployment process.
Reclaw fits naturally wherever agents create data that matters.
Flat files and JSON dumps become difficult to query, validate, share, and maintain as agent output grows. Reclaw keeps the local simplicity while adding real structure.
Hosted DBs add setup complexity and hide structure behind layers natural for human developers, not agents. Reclaw is optimized for how agents actually inspect and manipulate local environments.
An agent that cannot persist useful state is forced to repeat work, lose context, or depend on brittle workarounds. Reclaw gives agents a durable, reusable memory layer.
Structured records stored in a lightweight, fast, reliable database model.
Continuous protection with cloud backup for durability and peace of mind.
Collaborate across your team with shared access to agent-created records.
Each table has a clearly named file, making schema discovery intuitive for humans and agents.
Lightweight setup: install the skill and start persisting data.
Reclaw is built for teams using AI agents to do real work.
Using coding agents day-to-day.
Building agent-first products.
Wiring agents into operations.
Automating workflows with agents.
Managing structured outputs.
Building agent-assisted systems.
If your agents create data that matters, Reclaw gives that data a proper home.
Install Reclaw and give your agents a real system of record. If you are already using agent tools like OpenAI Codex or Claude Code, adding Reclaw is straightforward.
You do not need to redesign your stack to use Reclaw. Start with one agent workflow, one local workspace, or one team process. As your agent usage grows, Reclaw grows with it.
Without Reclaw, agent-generated data is scattered across prompts, scratchpads, logs, and temporary files. With Reclaw, that data becomes organized, persistent, backed up, shareable, and easy for agents to understand.
Install ReclawReclaw helps teams move from one-off agent experiments to reliable agent-powered systems.
Reclaw is a database built specifically for AI agents. It lets agents persist structured data locally, back it up automatically to S3, and share it with a team.
Reclaw is designed for coding and task-oriented AI agents such as OpenAI Codex, Claude Code, and similar agent environments.
Data is stored locally in a SQLite-style database, giving agents fast access from the local environment where they work.
Yes. Reclaw automatically backs up data to S3 for durability and recovery.
Yes. Reclaw supports team sharing with an experience similar to Dropbox, making it easy to collaborate around structured agent-generated data.
Reclaw uses a file-per-table structure with clear names, so agents can discover and understand the schema by exploring the local filesystem.
No. Reclaw is designed to be easy to adopt. Installation is done by installing a skill.
Reclaw is the missing data layer for agent workflows: local-first, automatically backed up, team-shareable, and easy for agents to understand. Stop losing valuable agent output in temporary context and scattered files.
Install Reclaw and turn your agents into systems that remember.
Get started