Notes from the AgentConnect
How we build AgentConnect, what we learn running agents in real team channels, and the occasional release note.
AgentConnect vs. Claude Tag: comparing the two ways to bring agents into team chat
Claude Tag brings a shared Claude into each Slack channel, with identities scoped by administrators. AgentConnect brings a roster of agents that can call one another, on any ACP-compatible runtime, into Slack, Telegram, Discord, Lark, and your GitHub, GitLab, Gitea, and Linear workflows — open source and self-hostable. Here's how the two actually compare.
Introducing Linear support
Assign Linear issues to your agents, follow their plans and tool activity, and keep results and status updates in the issue.
Introducing Gitea support
Connect AgentConnect to gitea.com or a self-hosted Gitea instance. Agents can respond to issues, review pull requests, and post their results back to the repository.
AgentConnect — the open-source, multi-agent alternative to Claude Tag
Bring multiple ACP-compatible agents into the conversations where your team already works with an open-source, self-hostable alternative to Claude Tag.
Grep beats LSP? Why coding agents ignore your fancier tools
In my study of lexical search and LSP-backed semantic navigation, I found that a tool's LLM-friendliness depends on more than retrieval precision: output context, task structure, and the surrounding harness also shape coding agent performance.
Why we built AgentConnect
Everyone on our team was building the same agent glue — message channels, cron jobs, credential handling, context stitching. One day we compared notes and found we were all writing nearly identical code. So we built it once.
Put an agent in your first channel.
Open source and self-hosted — run AgentConnect on your own infrastructure, bring any agent, and keep full control. No lock-in.
AgentConnect