Nakama

Nakama is an open-source AI agent platform built for teams, released under the MIT License by ahmadrosid. It provides a ChatGPT-style dashboard where each agent has its own identity, memory, instructions, tools, and permissions — all running on a single server that supports multiple organizations. The project is actively maintained on GitHub, with the latest release at v0.4.10 as of September 2026.

What It Is

Nakama is a self-hostable multi-agent platform that lets teams build, run, and manage AI agents collaboratively. It sits in the same category as Hermes Agent and OpenClaw — the GitHub description explicitly positions it as "like Hermes Agent & OpenClaw but designed to work nicely with teams" — but focuses on reducing setup complexity and enabling multi-org, multi-user deployments. Agents can use tools, skills, memory, and external channels, and the whole system can run from a single Docker container or through Nakama's managed cloud offering.

Architecture and Deployment Model

Nakama is written primarily in TypeScript and uses Bun as its runtime. It can be deployed in three ways:

• Local development: bun run dev:web starts both the server and web dashboard at localhost:3003.
• Docker: A prebuilt image is available at ghcr.io/ahmadrosid/nakama:latest, exposing the dashboard at port 4310.
• Managed cloud: Nakama Cloud at getnakama.cloud provisions a hosted instance with no need for Bun, Docker, or a VPS.

The server stores configuration in ~/.nakama/config.ini and exposes interactive API docs at /docs. A live demo is available at demo.getnakama.cloud.

Multi-Team and Agent Isolation

Nakama is designed from the ground up for multi-tenant use. A single server instance can host multiple organizations, each with its own members, agents, and isolated data. Each agent gets its own workspace, memory, and role — preventing cross-agent data leakage. Invites and role-based permissions are built in, making it suitable for teams that need governance over who can interact with which agents.

Integrations and Channels

Agents in Nakama can reach users across multiple channels and connect to external services:

• Messaging: Telegram, WhatsApp, and Discord are supported natively.
• External apps: Composio integration provides access to more than 1,000 external applications, enabled through the web UI under Integrations.
• Local models: Nakama supports Ollama for local inference, requiring no API key or external payment.
• Omni: The Docker image ships with omni, a tool that compresses bash and read_file output before it reaches the model, reducing token usage. It can be toggled off via the Integrations panel or the NAKAMA_OMNI=0 environment variable.

Skills, Scheduling, and Automation

Beyond one-off conversations, Nakama supports recurring scheduled work — agents can run jobs on a defined schedule (e.g., every Monday) without manual intervention. Agents can also convert successful workflows into reusable skills, enabling a form of continuous improvement. The homepage lists concrete use cases the project team says users are already running: building AI product backends, managing social media pipelines, writing books with multiple collaborating agents, automating daily reports, and turning podcasts into short video clips.

Update: v0.4.10

The latest release is v0.4.10, published on September 8, 2026. The repository was last pushed to on September 11, 2026, indicating active development. The project has accumulated 368 stars and 65 forks since its creation in May 2026, with 28 open issues. An ARCHITECTURE.md file in the repository documents the system design for contributors and self-hosters.