agor

Agor is an open-source platform designed as a team command center for agentic workflows, providing a shared workspace where human users and AI teammates collaborate in real time. It allows teams to manage coding agents and persistent AI helpers, supporting live presence features such as visible cursors, comments, and reactions on a Figma-like canvas. The platform emphasizes multiplayer collaboration, enabling engineers, reviewers, product managers, and QA professionals to work together on shared development environments and branches, eliminating the need for individual setups.

Key features include spatial boards for organizing branches, zones, sessions, and teammates; rich agent sessions that display tool calls, decisions, session trees, and handoffs; and a persistent knowledge base where both humans and AI agents can store decisions, runbooks, prompts, and reusable context. Agor supports the integration of various agent harnesses, including Claude Code, Codex, Gemini, Copilot, and OpenCode, allowing teams to select the most suitable provider per session without being locked into a single ecosystem. Agents can be given durable identities, memory, skills, and schedules, and can be taught through conversational onboarding, making it possible to standardize successful prompting patterns as reusable triggers.

The platform offers a message gateway for bringing agents into communication channels like Slack and GitHub, and includes a scheduler for automating workflows such as standups, audits, reports, and digests. Artifacts can be rendered directly on the board, including dashboards and mockups, and branch environments can be managed with start, stop, health-check, and log inspection capabilities. Agor also provides Unix-level isolation modes for enhanced security when required.

1 license, with a managed hosting option (Agor Cloud) in development for teams preferring not to self-manage infrastructure. The platform is built to integrate with existing MCP servers and allows both users and agents to perform actions via the MCP-native control layer. Its collaborative approach is designed to help teams move from solo agent experiments to visible, team-wide agentic workflows anchored on git branches.