AgentConnect
Open-source team agent platform for Slack, GitHub and more that asks Jev once per session which agent runtime and model to use.
Overview
AgentConnect is an open-source platform where teams and multiple AI agents work together in Slack, Telegram, Discord, Lark, Google Chat, GitHub, GitLab, Gitea and Linear. You bring Claude Code, Codex, Grok Build, DeepSeek, Pi or any ACP-compatible agent into shared conversations, give each agent a role, and let people and agents call each other, remember what they learn, and start work from messages, issues, pull requests, webhooks or schedules.
Jev is used for agent routing and model selection. Per the list summary, it asks Jev a typed Choice once at session start, not per message, to pick which coding-agent runtime and model runs the session, chains follow-up decisions and pins the winner, with a 5-second deadline that falls back to the agent's configured runtime. The README also mentions using Jev to route new support conversations to the right specialists.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which coding-agent runtime and model should run this session? | — |
| Choice | Which specialist should this support conversation go to?(inferred) | — |
How it works
- A message, issue, PR, webhook or schedule starts a session
- Jev picks runtime and model once, within a 5-second deadline
- On timeout, fall back to the agent's configured runtime
- The pinned agent works in the shared thread with the team
Strengths
- One decision per session keeps cost low
- Deadline with fallback avoids stalls
- Works across many chat and code platforms
Caveats
- Large platform; Jev is one optional part
- Routing needs TypeSafe access
At a glance
- Type
- Service
- Status
- beta
- Language
- TypeScript
- Runtime
- self-hosted service / cloud
- Has code
- yes
- License
- Apache-2.0
Use cases and integrations
How curated lists describe it
Team agent platform: asks Jev a typed `Choice` once at session start, not per message, to pick which coding-agent runtime and model runs the session, then chains follow-up decisions and pins the winner behind a 5-second deadline that falls back to the agent's configured runtime.
yibie/awesome-jev
Frequently asked questions
How often is Jev called?
Once at session start, per the listing, not on every message.
What if Jev is slow?
After a 5-second deadline it falls back to the agent's configured runtime.
Which agents can join?
Claude Code, Codex, Grok Build, DeepSeek, Pi or any ACP-compatible agent.
Related projects
Three Spanish-language demos where Jev decides and Claude reasons: Jira triage, code pre-flight and a PR review router.
AppprototypeTypeScriptAlibaba's AgentScope 2.0 agent framework, which added a JevClassifierModel for binary, choice and score decisions via the TypeSafe SDK.
ProjectproductionPython- jev-router★ 593
Per-turn model routing for Claude Code and Codex: wraps the real CLI and lets Jev send simple turns to fast models and hard ones to strong.
ProxybetaJavaScript Glue that turns one Linux machine into a team of coding agents, with Jev making typed calls like who builds a task and whether it may merge.
ProjectbetaTypeScriptTypeScript/Bun launcher that asks Jev which model and reasoning effort each Codex, Claude Code or Pi turn should use.
ProjectbetaTypeScriptDesktop workspace for coding agents that runs tasks in parallel worktrees, listed as using Jev to pick the best agent per task.
ProjectbetaTypeScript, Rust