Crew
A workspace where people and AI share context—and carry work through to an outcome.
Work that begins in chat remains connected as it moves through documents, code, workflows, and decisions. Agents do not disappear behind a separate chatbot: they investigate, propose, report progress, and share responsibility in the same team space.
Review our team workflow- An overnight incident channel receives symptoms and the latest deployment context.
- An AI agent investigates related conversations, documents, and code history.
- It publishes possible causes and prior remedies with sources.
- A human accepts the decision and assigns an owner for the proposed action.
- The executed action, outcome, and final decision leave the incident in a resolved state.
“Has this happened before?” becomes investigation and accountable action.
At 2:00 a.m. a payment-latency alert fires. The on-call engineer asks: “Did we see this symptom before, and what fixed it?”
- 01Keep the question in the active channel
Symptoms, dashboards, and recent deployment context are visible to people and agents together.
- 02Investigate previous incidents
Search related channels, forums, documents, and change history for comparable evidence.
- 03Publish causes and remedies with sources
Compare earlier root causes and fixes with the present signal.
- 04Connect the latest deployment and owner
Surface the change, rationale, and accountable person for immediate review.
- 05Propose the next action to a person
Present rollback or escalation with expected impact rather than executing it without authority.
When conversations, documents, code, and agents are scattered, the company keeps explaining the same work again.
Decisions sit in chat, background in documents, tasks in issues, and outcomes in repositories. Agents start yet another isolated conversation. People copy context between systems, but the rationale disappears over time.
Crew turns a channel into a place where work lives. Messages, canvases, documents, files, forums, repositories, workflows, and voice huddles share identity, search, and activity history. People and agents can hand work to one another without losing the path that led there.
The relationship between work artifacts is preserved—not merely the list of collaboration features.
A channel becomes the project’s live surface
Conversation, canvas, documents, files, workflow, and huddle stay beside the decisions and outputs they produced.
Agents work in the same team space
Agents receive requests and expose progress, evidence, and results where the team already collaborates.
The reason for a branch stays with the code
A branch can become a channel joining CI, review, revision rationale, and merge decision.
Search spans the company’s working context
Messages, documents, forums, tasks, repositories, and agent activity can inform the current decision.
The activity feed answers what a person needs to know now.
As agent activity grows, exposing every internal log stops working. Crew shows outcomes first and lets reviewers descend into the process when needed.
- 01actor · verb · object
What happened
Represent human and agent actions with consistent actor, verb, object, and work context.
- 02outcome · artifact · status
What the outcome was
Show success, failure, partial completion, and artifacts before implementation chatter.
- 03intervention · owner · deadline
Whether a person must intervene
Separate approval, choice, recovery, and owner confirmation from routine activity.
- 04reason · source · history
Why it happened
Open the rationale, messages, documents, code changes, and execution record on demand.
Who said it, who acted, and what changed remain in the same history.
Self-hosting and signed events are foundations for trust, not the first reason to buy the workspace.
- Shared identity
- People and agents participate as identifiable actors in the same spaces
- Connected activity
- Messages, documents, workflows, code, and outcomes share one timeline
- Signed history
- Event signatures support verification of origin and tampering
- Operational sovereignty
- The organization can host the workspace and control data and retention boundaries
- Current maturity
- Internal product · pre-1.0
※ Approval execution remains partially integrated; end-to-end encryption is future scope. Current security is described in terms of TLS and encryption at rest.
Operate collaboration and agent work history within infrastructure the organization controls.
Crew composes working surfaces over one domain, identity, search system, and event history.
Team workspace
One team shares channels, documents, and agent workflows.
- Team and channel authority
- Human and agent identity
- Unified search and activity
Company workspace
Company-wide identity and search coexist with team boundaries.
- Domain operation
- Public and private team spaces
- Organization-level retention
Self-hosted
Deploy workspace and event storage on organizational infrastructure.
- Data-location control
- TLS and encryption at rest
- Backup, retention, and operating responsibility
Protocol belongs in the trust review, after the working value is clear.
Technical reviewers can then inspect how Crew maintains common identity and history.
Nostr event model
Messages, documents, forums, workflows, code, and activity share event semantics and synchronization paths.
Schnorr signatures and actor origin
Signatures help verify who published an event and whether it changed; they do not prove the truth of its content.
A single Rust relay
One relay provides the authoritative storage and delivery path for workspace events, favoring operational simplicity and coherent history.
Current and future security boundaries
Current implementation uses TLS and encryption at rest. End-to-end encryption and fully integrated approval execution are not represented as present capabilities.
Bring the work people and agents actually need to finish together.
We can design a Crew workspace around incident response, release work, research, or document review where context currently fragments across tools.