We ship continuously. This changelog highlights user-facing improvements and the reasoning behind them.
Follow along
- Twitter: https://twitter.com/filepad_ai
Recent updates
Agent Access API v1
External agents can now connect to Filepad workspaces via a public HMAC-authenticated API.
- Environment read — agents inspect canonical folders, file tree, and supported file content
- Skills as prompts —
skills/*.mdfiles are discoverable as callable capabilities via/prompts - Safe artifact creation — agents write to
artifacts/without a review step - Proposals — writes to
memory/and other sensitive paths are staged as diffs for human review - Activity events — agents push structured events to the workspace activity log
- Per-workspace scoped keys — keys cannot cross workspace boundaries
- Key lifecycle — create, rotate secrets, patch scopes, and revoke without downtime
See the Agent Access API v1 reference for full documentation.
Review and approval UI
The Reviews panel now shows pending proposals from connected agents. Review additions and removals line-by-line, accept or reject with a single click. Every decision is timestamped and attributed to the reviewing user.
Agent Access key management
Workspace admins can create, rotate, and revoke Agent Access keys directly from the Agents panel. Scopes are adjustable without re-issuing the key.
Standard environment folders
All new workspaces initialize with the canonical 8-folder structure: .filepad/, agents/, skills/, memory/, sources/, uploads/, artifacts/, automations/. Every file is real, editable, and API-accessible.
FilepadAI
FilepadAI is the native in-app assistant. Ask it to explain access, summarize file contents, debug unexpected behavior, and propose reviewable improvements through the normal Filepad run and approval system.