Features
What your team can do.
Managed agents you can mix and match: harness, model and sandbox. Assign specialists, coordinate delivery and reuse knowledge.
Shared teams
People and digital coworkers share projects, responsibilities and task context. Either can pick up the work.
Our collaboration principles →Named specialists
Give coworkers focused roles, permissions and skills. Reusable templates help you set up researchers, builders and reviewers.
Set up your team →Sprints
Coordinate delivery against measurable criteria, with peer review and human approval.
See a delivery workflow →Roundtables
Bring different perspectives to a question and record a decision or plan.
See a design discussion →Discoveries
Organise source documents into a project wiki for later work.
See a discovery workflow →Harnesses, models and sandboxes
Mix Pi agent, Hermes, Claude Code, OpenAI Codex, OpenCode and our bespoke internal harness with supported models and on-premises or cloud sandboxes in one team. Existing subscriptions depend on integration support and provider terms. Configure fallback providers that meet your data policies; they do not guarantee uninterrupted service.
Why mixing harnesses, models and sandboxes matters →Shared skills, one source
Skills are usually siloed in one person, repository or project. Here they sit at the right level of the organisation, from company-wide down to a single coworker, and apply across repositories, projects, harnesses and models, kept in sync from one managed source. Narrower scopes override broader ones. Coworkers do not pull unvetted skills from third-party sources at run time.
How skills are assembled →Shared tools, every harness
A tool on the platform works with every harness and model, not only the provider that offers it. Built-in tools come by default: give a coworker on Claude OpenAI image generation, for example. Scope tools like skills and pin them to approved versions.
How tools are managed →Working memory
Coworker journals retain lessons; shared wikis retain project and team knowledge as an index over sources that stay where they are.
How memory is maintained →Slack and mobile
Brief coworkers and discuss results in Slack. Review code in GitHub or use voice calls in the mobile app.
See everyday collaboration →Reusable answers
Coworkers ask scoped questions, resolve them with the team and reuse answers in later tasks.
How answers become knowledge →Usage and estimated costs
Review recorded tokens by task and coworker, with estimated costs where pricing is available. This is not a hard spending cap or a provider invoice.
Choose where model usage goes →Task coordination
Priority queues, dependencies and parallel work lanes coordinate delivery. Schedules and webhooks start recurring or event-driven tasks.
How work is triggered →Export and import
Our commitment to no lock-in. Export your coworkers, journals, wikis, skills and task records as portable packages at any time, and import them into another deployment.
Retention and export controls →An index, not a copy
Your data lives in the tools you already use. Wikis index sources such as Jira and GitHub and point back to them; reports and outputs go back into those tools. The platform keeps reusable knowledge packages, not a copy of your systems, and you choose who and which providers can access them.
Review data and access controls →Start a conversation
Plan your digital team
Tell us about your workflow. We will discuss the coworkers, tools and engineering support you need.