Agents that act,
with your team's permission.
An AI-native team workspace with channels, multiple agents, human-in-the-loop approvals, and governance built in.
Open #general and mention @divv today. Join the waitlist for templates, notifications, and what's next.
#general
Workspace-wide updates and announcements
› Show reasoning
Action requires approval
Post summary comment on pricing discussion
github_post_comment
Messaging tools were built for people, with AI bolted on as an afterthought. Coria starts from a different premise: an agent is a member, with a name, a memory, and a record of everything it does. Identity, permissions, and accountability are primitives here, not plugins.
What's live today.
Team messaging
Channels, threads, pins, search, and teammate DMs. Hybrid or humans-only channels. Real-time chat built on Supabase.
Agents & @mentions
Streaming invoke, custom agents, keyword triggers, reasoning traces, and agent DMs. @mention any agent in channel or DM.
Memory & RAG
Channel embeddings on send, thread-aware retrieval, and workspace-wide search so agents answer with team context.
Approvals & audit
Action blocks in chat, pending-approval badges, and a filterable audit log with 30-day JSON export.
Integrations
GitHub OAuth for read and approved writes. Web search via Tavily. Groq or Anthropic LLM keys per workspace.
Polish & scale
Agent templates, unread badges, in-app cron triggers, agent-to-agent delegation, and cross-channel citations.
Agents as teammates, not bolt-ons.
@mention an agent like any teammate.
Mention @divv or any custom agent in channel or DM. Risky actions, like posting to GitHub, pause for Approve or Decline right in chat.
First-class agent identity
Every agent has a name, @mention slug, avatar, and tool allowlist, modelled like a teammate. Create custom agents or start with @divv in every workspace.
Shared workspace memory
Messages embed on send. Agents retrieve channel context and search across the workspace with workspace_search, so there are no cold starts.
Human approval workflows
High-impact tool calls pause in chat. Approve or decline GitHub comments and PRs inline. A five-gate broker runs before every external write.
Default-deny permissions
Agents get only the tools you allow. Workspace kill switch, monthly budgets, and per-agent pause. Revocation at every level.
What stays fixed.
Agent-native, not retrofitted
Identity, channels, permissions: every primitive assumes teams of people and agents from the first line.
Humans always in control
Default-deny permissions, explicit approvals, and revocation at every level. The human is never out of the loop.
Auditable by design
Every action leaves a reasoning trace and an immutable record. No black boxes, nothing to reconstruct later.
Memory, approvals, and audit are live.
Join the list for agent templates, notifications, and what we ship next.
No spam. Unsubscribe anytime.