Effective collaboration between AI agents requires a communication layer where they can assign work, share results and track progress—capabilities that Vercel is now providing through Board, an experimental shared message platform.

Vercel built Board initially for its own security operations, using it to coordinate agents investigating vulnerabilities and preparing fixes. The company subsequently restructured v0's software factory around Board, which now merges over 50% of PRs, according to Vercel's Wednesday announcement.

It packages what we've learned about coordinating work across agent fleets and builds towards a future where multiple teams of humans and agents orchestrate swarms, software factories, and other agentic workflows.

Vercel

The coordination conundrum

Multi-agent communication has proven essential for solving complex problems at scale. OpenAI demonstrated this in September when it announced that agents had resolved the Navier–Stokes problem, a long-standing mathematical puzzle about fluid dynamics. The effort involved roughly 10,000 concurrent agents that exchanged information throughout the process.

OpenAI reported that 2.7 million messages were sent and approximately 130 billion output tokens were consumed during the Navier–Stokes breakthrough.

Software development is increasingly requiring parallel agent orchestration. Last week, Google and Cursor each unveiled competing strategies for coordinating coding agents—Google targeting single-developer scenarios while Cursor enables cross-team agent communication. Replit introduced Missions on Tuesday to bring multiple agents together on larger projects.

Beyond purpose-built platforms, organizations are repurposing existing collaboration tools. Cognition launched Devin Bots on Tuesday, leveraging Slack to connect agents with human teammates. Salesforce CEO Marc Benioff stated in an X post Thursday that more than 226,000 custom agents have been deployed to Slack across major agent-building platforms, noting this represented an unanticipated use case when Salesforce acquired Slack in 2021. Benioff characterized Slack as now serving as trusted agent infrastructure.

Vercel added its Agent to Slack code channels in August, enabling team members to observe agent activities, provide direction and review proposed changes.

Board provides channels and threaded conversations where agents can claim work, share discoveries, monitor task dependencies and log completed activities. The platform maintains a complete activity history, enabling developers to review agent behavior.

A key design principle is Board's technology-agnostic architecture. Vercel explicitly supports Claude Code, Codex, Pi, OpenCode, eve, fx, Mastra and LangChain, allowing tools built on different technologies to collaborate through Board.

This is an important building block for software factories, self-optimizing systems, and scientific research.

Guillermo Rauch, Vercel CEO

Vercel CEO Guillermo Rauch characterized Board as a message board infrastructure primitive or headless Slack for agents. He emphasized that Board operates independently of any particular model, agent framework or hosting infrastructure.

Just bring your agents, give them our CLI, and watch magic happen.

Guillermo Rauch

Keeping humans in the loop

Board integrates with tools developers already rely on, including Slack, GitHub and Linear. A developer can submit a request via Slack, which an agent picks up and decomposes into smaller tasks for other agents. Board maintains the resulting discussion, task assignments and progress updates.

As work advances, Board synchronizes with GitHub, allowing agents and developers to monitor pull request changes and review comments without manually transferring information between systems. Event-driven triggers enable agents to resume work when new messages arrive or task statuses shift.

Human oversight remains central to the design. Agents can report findings and await developer approval before proceeding. Board also implements channel-level permissions controlling which people and agents can access information or take specific actions.

The security question

Independent agent communication raises critical concerns about control, access and information sharing. Rauch cited the OpenAI and Hugging Face cybersecurity incident from July as a high-profile case where agents used an internal software repository to exchange findings and coordinate activities.

A core part of why we built Vercel Board was to fulfill our own needs for security and auditability.

Vercel

Security and oversight are fundamental to Board's design. Vercel implemented access controls and activity logs, enabling teams to limit which agents can interact with specific channels and audit their actions.

Vercel's documentation cautions that agents may read messages as instructions, making it essential to control who can post in channels where agents receive assignments. By default, channels are open to team members, project deployments and installed apps unless administrators impose restrictions. Vercel recommends establishing private channels or tightening permissions where needed, and keeping sensitive data such as API keys and tokens off Board entirely.

Board is available now as an experimental feature across all Vercel plans, accessible via the Vercel CLI or JavaScript SDK. The service is free during the experimental phase, though Vercel indicated it may introduce pricing later.