Welcome to Seshat - building the future of agentic AI together 🚀

[EngineerProjects]

Hello, and welcome 👋

Seshat is now open to the world. Whether you are a developer, a researcher, or just curious about where agentic AI is going, you are in the right place.

---

What is Seshat?

Seshat is an open-source Go runtime for autonomous AI agents. It ships as a single binary with no external dependencies: persistent sessions, 60+ sandboxed tools, 15 LLM providers, MCP client, RAG, a skills system, a gRPC API, and a full terminal UI.

But the binary is only the starting point.

---

The bigger picture

The agentic AI space is exploding almost entirely in Python. Seshat is one of the very few production-grade Go runtimes in this space. A single binary, low latency, easy deployment, no dependency hell.

More importantly, we believe the real unsolved problem is not what a single agent can do it is how multiple agents can work together.

Most systems today use pipelines and orchestration graphs. These are useful, but they don't model work the way humans actually do it.

Our thesis: the right model for agent collaboration is the human organization. Real teams have roles, inboxes, meetings, status reports, shared decisions, and a mission memory. We want to reproduce those patterns with autonomous agents not as a metaphor, but as actual primitives in the runtime.

---

What we are building toward

• ✅ Today: solid single-agent runtime (CLI, TUI, SDK, gRPC)
• 🔨 Next: team primitives agent mailboxes, roles, mission memory, structured inter-agent communication
• 🔭 Vision: a unified environment where humans and agent teams collaborate on real work

---

How to get involved

This is an open research and engineering project. We do not have all the answers and we genuinely want input.

• 💡 Ideas → open a discussion in the Ideas category
• 🐛 Bugs / issues → GitHub Issues
• 🔧 Code contributions → read CONTRIBUTING.md and open a PR
• 💬 Questions → Q&A

We are building this together. Thank you for being here. 🙏