Thenovi AI Ltd., a company developing an interaction infrastructure for artificial intelligence agents, today announced the launch of its developer platform that connects coding agents for shared conversations.

As more coders adopt AI agents, they have begun to specialize them into different roles: planning, review, integration and testing. This is because, in many cases, developers discovered certain models have different quirks that make them more efficient at particular tasks.

For example, a developer might use Claude Code to orchestrate a workflow and provide high-level planning and architecture, Codex for code implementation, and a fleet of other agents for testing and review. “Coding is becoming agent-based, but the agents themselves can’t actually work together,” said co-founder and Chief Executive Arick Goomanovsky.

Goomanovsky pointed out that after each iteration, it comes down to the developer to take the output from one agent and then feed it into the next one. This becomes a tedious dance of copy/paste or document shuffling between tools.

Thenovi seeks to resolve this by becoming a chat-hookup backend that can spin up agents, providing them with a shared space where the developer can communicate with them as if they were a team of employees.

The software engineer can then provide the plan document to the planner, which then crunches on the schema to get the architecture of the project polished. Once that’s ready, it can be passed off for the reviewer to fill in any gaps before implementation, which can request the planner to fix up the plan document.

Next, the finalized project requirements get handed off to a coding implementation agent, which can break it down into subtasks and coding agents get to work. Finally, testing and security agents get the finishing touches squared away, which could iterate back to implementation for code fixes.

“Instead of wiring agents together through files or local shared state, developers can connect agents to a common environment where coordination happens automatically,” explained Chief Technology Officer Vlad Luzin.

According to Thenovi, context, files and workflow remain coherent the entire time as every agent retains persistent knowledge of its role and last action within the project. Under the hood, orchestration contains specialized logic to prevent rationale loops, dead-end pathways and overthinking.

The launch also arrives as personal AI agents, such as OpenClaw, have begun to emerge in the open-source community and service providers. These agents act as colleagues that can reason, plan and take action for their users. However, like many current enterprise-level agents, they operate in isolation. They also lack a secure way to discover and connect with other agents without additional plugins.

Thenovi said its platform can extend beyond development work through what it calls Agentic Mesh, a connectivity layer for independent agents to discover each other securely. Using the mesh, agents can securely connect, communicate and delegate tasks within user-set boundaries.

With the platform, a personal agent could coordinate directly with another agent or interact with a vendor’s service agent to negotiate tasks, coordinate logistics or execute transactions while maintaining identity controls.

Images: SiliconANGLE/Microsoft Designer, Thenovi