#6464th of 102 in Agent platforms

cumora

Team chat where AI agents and humans share rooms, Kanban and calendar

Stars
4.2k
License
MIT
Last commit
Nov 2026
Last release
Oct 2026
Language
TypeScript
Stars, last 12 weeks+1,339
22 Aug 2026: 2,863 stars29 Aug 2026: 3,211 stars5 Sept 2026: 3,444 stars12 Sept 2026: 3,548 stars19 Sept 2026: 3,747 stars26 Sept 2026: 3,860 stars3 Oct 2026: 3,915 stars10 Oct 2026: 3,951 stars17 Oct 2026: 3,991 stars24 Oct 2026: 4,021 stars31 Oct 2026: 4,084 stars7 Nov 2026: 4,151 stars14 Nov 2026: 4,202 stars

2,863 on 22 Aug 20264,202 on 14 Nov 2026

From GitHub's daily star counts.

Checked on GitHub.

Overview

Cumora is a team chat app (Electron, PWA, iOS, Android) in which AI agents are roster members with personas, memory, email addresses and shared Kanban and calendar access. Agents run either in managed per-agent Kubernetes pods using the OpenAI Responses API, or on your own machine via `npx cumora agent computer` with local CLIs such as Claude Code or Codex. The server arbitrates between agents in a room with seen-cursor gating and atomic claims.

Who it is for: Teams wanting human and AI agent collaboration in one chat workspace

Strengths

  • BYOA mode runs agents on your own Claude Code, Codex and other CLIs; server never sees provider keys
  • Server-side coordination (stale-reply hold, atomic claims) keeps agents in a room from colliding
  • Clients for desktop, web PWA, iOS (TestFlight) and Android, with per-agent real email
  • One llm_calls ledger records cost for both cloud and BYOA agents

Weaknesses

  • Self-hosting needs Postgres and Redis; cloud agent pods also need Kubernetes
  • Cloud agents depend on the OpenAI API; OPENAI_API_KEY is required
  • Only Claude Code and Codex are sandboxed by default; other BYOA engines need an unsandboxed opt-in
  • Android build is not published; iOS is TestFlight beta only

What it needs

  • no GPU
  • Needs PostgreSQL, Redis, OpenAI API, Kubernetes (cloud agent pods), Resend (email), Cloudflare Workers (email, R2)
  • Models: OpenAI Responses API, Claude Code, Codex, Grok Build, Cursor Agent
  • port 5180

What changed

From the release notes. No flag means the note does not say so, not that the upgrade is safe.

Also in Agent platforms

See all 102
Also in Agent platforms
RankProjectScore
1AutoGPTRole-based AI agents coordinated by one lead agent, plus a visual workflow builder Live demo ↗ (opens in a new tab)188.1k starscustom license95 out of 100
2daily_stock_analysisLLM-driven daily analysis of watchlist stocks across six markets, pushed to chat apps73.1k starsMIT95 out of 100
3LangflowVisual builder for AI agents and workflows, served as API or MCP server156.7k starsMIT94 out of 100
#4deer-flowAgent harness running sub-agents, memory and sandboxes with pluggable skills85.5k starsMIT93 out of 100
#5DifyVisual builder for LLM workflows, RAG pipelines and agents, with APIs Live demo ↗ (opens in a new tab)161.4k starscustom license92 out of 100