I built WritBase because AI agents need a shared, persistent task registry - not ephemeral state that vanishes between sessions.
It's an MCP server that gives your agent fleet:
- Scoped permissions (6 types: read, create, update, assign, comment, archive) per project and department
- Full provenance — every change logged: who, what, when, why
- Inter-agent delegation with depth limits and cycle detection
- Optimistic concurrency control so agents don't silently overwrite each other
- Webhook delivery (HMAC-signed, Standard Webhooks spec)
Stack: Supabase (Postgres + Edge Functions) + Next.js dashboard.
Deploy in 3 commands on the free tier.
Apache 2.0.
Any MCP client works - Claude Code, Cursor, VS Code, Windsurf.
Agents connect, create tasks, update status, delegate to each other.
Would love feedback on the permission model - balancing security (agents shouldn't touch what they can't) vs. usability (agents shouldn't need admin help for every task).
RusDyn•2h ago
It's an MCP server that gives your agent fleet:
- Scoped permissions (6 types: read, create, update, assign, comment, archive) per project and department - Full provenance — every change logged: who, what, when, why - Inter-agent delegation with depth limits and cycle detection - Optimistic concurrency control so agents don't silently overwrite each other - Webhook delivery (HMAC-signed, Standard Webhooks spec)
Stack: Supabase (Postgres + Edge Functions) + Next.js dashboard. Deploy in 3 commands on the free tier. Apache 2.0.
Any MCP client works - Claude Code, Cursor, VS Code, Windsurf. Agents connect, create tasks, update status, delegate to each other.
Would love feedback on the permission model - balancing security (agents shouldn't touch what they can't) vs. usability (agents shouldn't need admin help for every task).
GitHub: https://github.com/Writbase/writbase