Services
- Product Strategy
- Interface Design
- AI Orchestration
- Integrations
- Application Engineering
Technology Stack
- React
- TypeScript
- Next.js
- Supabase
- PostgreSQL
- Trigger.dev
The challenge: keeping shared work connected
At Codeberry, eight people can form several overlapping teams, each working on a different part of a project. Code, documentation, planning, and conversations live across tools such as GitHub, Linear, Notion, and Slack.
Keeping everyone informed meant moving between those tools and carrying context from one place to another. Introducing AI added another challenge: how could agents participate in shared work without creating more isolated conversations and more copying and pasting?
We wanted a place where people could communicate, find relevant information, and work with agents dedicated to specific tasks.
Our approach: build around the conversation
We designed Omni around shared conversations between people and AI agents. Teams can bring in their documentation, configure agents with specific instructions and skills, and connect the tools their work depends on.
The aim is to make information and assistance available where the discussion happens, while keeping people involved in decisions and approvals.
People and agents in the same conversation
A chat in Omni can include any mix of people and agents. Anyone can bring an agent in with an @mention, and the agent replies in the thread with the context of the discussion so far.
Agents ask before they change anything. When an agent wants to write a file, people in the chat can approve or decline the action. Replies an agent gives someone privately stay private unless they choose to share them.
The same pattern works whether a team is researching a market, scoping a feature, or tracking a launch.
Agents with a defined role
Each agent has a name, a handle, instructions, and a model. Teams can add tools, skills, and knowledge, and decide how the agent takes part in chats and whether it can change its own rules.
An instructions score suggests improvements, helping teams write instructions an agent can follow consistently.
From shared knowledge to action
Omni can help teams find information in documents and previous conversations, prepare outputs, and take action through connected tools.
Specialist agents can contribute to different parts of the work. People can review proposed actions, provide direction, and keep the work grounded in their goals.
Each person connects their own account to tools such as Linear, Notion, Jira, and Stripe, and an agent uses that connection only for what it is asked to do.
Built around Bring Your Own AI (BYOAI), Omni lets users connect their own credentials for supported AI providers and choose models that fit their needs.
A marketplace of agents and skills
Teams don’t have to start from a blank page. The marketplace offers ready-made agents for everyday, community, creative, and work tasks, along with teams of agents and reusable skills such as meeting notes, status updates, and document review.
Each agent can be installed as it is or customized, and comes with conversation starters to help people get going.
Getting started with OmniMate
Every workspace starts with OmniMate, a general assistant that helps people set up their work. Describe a goal, and OmniMate can suggest and create a specialist agent, then help invite the right people.
Agents can also take on work on a schedule, such as a weekday news summary or a weekly status update.
Built by the team that needs it
Codeberry designed and developed Omni across product strategy, interface design, AI orchestration, integrations, and application engineering.
Our own work provides an ongoing proving ground: a small team, multiple responsibilities, and a need to keep knowledge and communication connected.
Being explored across different kinds of work
Omni is currently in closed beta. Testing groups include software development teams, an arts management institution, a couple coordinating their house construction project, and a small nonprofit managing work around public grants and projects.
These groups bring different needs to the same underlying challenge: keeping people, information, and ongoing work connected.
What’s next
We are refining Omni through internal use and closed-beta testing, with a public BYOAI offering planned.
