Pre-release · Self-hosted
A swarm of agents for your infrastructure.
Tahyi coordinates specialist AI agents that run deployment, monitoring, observability, and database operations — autonomously on reversible work, gated on the rest.
$ git clone https://github.com/kay-g2/tahyi-site.git
$ cd tahyi-site && npm install
$ npm run dev
# → site at http://localhost:4321Quickstart
Self-hosted. Dry-run by default.
Clone the repo, register a non-production environment, and enable your first specialist agent — with every action traced in the audit log before it touches real infrastructure.
Why a swarm, not a chatbot.
01
Specialists, not generalists
Deployment, monitoring, observability, and DBA each get a narrow expert — not one model guessing across your entire stack.
02
Coordinated, not isolated
A shared layer routes tasks, context, and handoffs so agents act as a colony, not a pile of scripts.
03
Autonomous with safety
Reversible operations run on their own. Destructive and one-way-door actions hit approval gates — always.
From toil to traced automation.
Register your environment
Point Tahyi at a beachhead cluster or account. Start in non-production.
Enable specialists
Turn on the agents you need — deployment first, then monitoring, observability, DBA.
Review and run
Dry-run plans land in the audit log. Approve one-way doors. Let reversible work flow.
Everything you need to run infra with agents.
Domain experts
Four founding agents own deployment, monitoring, observability, and DBA — each replaceable without rewriting the swarm.
Shared operating layer
Tasks, context, handoffs, and heartbeats — an explicit contract between agents.
Two-way / one-way doors
Autonomy on reversible ops; gates on destructive and irreversible actions.
Full trace
Every plan, tool call, and outcome logged — append-only, replayable.
Ops without a full SRE org
Production-grade operations for teams that cannot staff 24/7 experts across every domain.
Self-hosted · MIT planned
Designed for your infrastructure. Public release pending board approval.
Bring your own models.
Tahyi connects to the LLM providers you already use. Configure credentials via environment variables — never commit secrets to the repo or agent prompts.
Built for production teams.
Tahyi is designed for self-hosted deployment with an MIT license at public release. Documentation is Markdown-first. Contributions open when the board approves public repos.
Frequently asked questions.
- How is Tahyi different from ChatGPT or a single coding agent for ops?
- Tahyi is a coordinated swarm of domain specialists with safety gates and an audit log — not one generalist session. Each agent owns a narrow slice of infrastructure operations.
- Will agents run destructive commands without asking?
- No. One-way-door actions require explicit approval. Dry-run is the default on early builds.
- What does Tahyi support at launch?
- 0.1 targets a documented beachhead environment. The four founding specialists — deployment, monitoring, observability, DBA — ship before breadth expands.
- Is Tahyi open source?
- Designed for self-hosted deployment with an MIT license at public release. Repos are private during pre-release — track progress on TAH-74.
Hire your first specialist.
Self-hosted. Dry-run by default. Every action traced.