AgentConnect today announced the release of its open-source platform designed to enable teams and multiple AI agents to work together across the communication and development tools they already use. The platform aims to address the limitations of current AI agent workflows, which often remain confined to individual terminals and fragmented contexts.
According to the company, AI agents are becoming increasingly capable, but most agent work still starts and ends with one person. Sessions typically live in separate terminals, context remains isolated, and teams build bespoke bots and integrations for each new workflow. Products that add a team layer often require moving work into another dedicated workspace or tying it to a single agent provider.
AgentConnect brings agents into existing team conversations instead. Teams can connect Claude Code, Codex, and other ACP-compatible agents, assign each a distinct role, and enable collaboration between people and agents through shared conversations and agent-to-agent calls. This approach allows agents to work alongside human teammates in the same platforms where work is already happening.
“AI agents become much more useful when they can work with the rest of the team, not just one person in one terminal,” the AgentConnect team said. “AgentConnect brings agents with different roles into the conversations where work is already happening.”
For example, a team responding to an incident can ask a triage agent to inspect the available context, bring in a code specialist to trace a suspected regression, and use a verification agent to check the proposed result. Human teammates can add context, redirect the work, and follow the diagnosis and decisions in the same Slack thread.
The same model applies beyond incident response. A support issue can move from a customer-facing Telegram conversation into a trusted Slack workspace without copying context by hand. Engineering teams can combine general and specialist reviewers around the risk of each pull request, with inline GitHub comments and independent check results. Scheduled and webhook-driven work can also continue in a shared conversation when human input is needed.
AgentConnect includes several key features:
- Multiple agents, one team: Configure distinct models, workspaces, memory, MCP servers, skills, repository access, and sandbox policies for each agent.
- Work across existing tools: Connect agents to Slack, Telegram, and Discord; start work from conversations, GitHub pull requests, webhooks, or schedules; and continue work between trusted messaging workspaces.
- Fine-grained access control: Manage agent access separately from session visibility, control which tools, repositories, and other agents each agent may reach, and link multiple social accounts to one member profile so supported platform permissions can be applied consistently.
- Open source and self-hosted: AgentConnect is licensed under Apache 2.0 and can be deployed with Docker Compose or Kubernetes. Agent execution and workspaces remain in the environment the team operates.
AgentConnect OSS is available today. Explore the source and star the project on GitHub, learn more at agentconnect.md, or read the documentation.
