AgentLab is releasing ProjectClaw, a Slack AI assistant built for teams that manage multiple projects. It answers the routine questions that consume a project lead’s day—“what’s open right now?”, “what shipped this week?”, “what did we decide about auth?”—by pulling together GitHub repo activity and Granola meeting notes.
Code + Meeting Context in One Place
Most project status is split between GitHub and meeting transcripts. ProjectClaw connects the two. A single Slack channel maps to a project with explicit repos and a Granola folder, so the bot can answer questions that span both sources and cite where the answer came from.
- Open work — non-draft PRs awaiting review, with repo and author
- Recent shipments — merged PRs in the last 24 hours or week
- Meeting summaries — latest Granola note for the project folder
- Decision search — recent meeting notes plus closed PR descriptions
- Ownership — cross-reference Granola action items with GitHub assignees
Answers cite their sources. Every response references PRs as repo#NUMBER or meetings by title and date. On tool failure the bot returns a partial answer and surfaces the failure inline—it never silently fabricates. View on GitHub →
Scheduled Standups
ProjectClaw posts two recurring updates per channel. A daily standup (Tuesday–Friday at 9am) covers open PRs and the last 24 hours of merges. A weekly summary (Monday at 9am) covers the last seven days of merges, aging open PRs, and issue activity. Both are driven by cron jobs registered in the Slack workspace and processed by the same agent loop that handles user mentions.
Scoped by Design
When a Slack workspace hosts many projects, a helpful bot can become a liability if it leaks context across channels. ProjectClaw maps each Slack channel to a specific project scope, and an always-on skill forbids tool calls outside that scope. If a user asks in an unmapped channel, the bot refuses to guess and asks which project to use.
Visit the ProjectClaw project page for the full architecture, setup guide, and stack.