About server-os
Server OS is an agentic operating system for teams that need to run agents as managed processes. It combines cost control, governance, observability, a web control plane, and a production runtime with offline mock mode.
Key features
- 01Agents as managed processes
- 02Cost controls and governance
- 03Observability
- 04Web control plane
- 05REST API, CLI, and SDK
- 06Offline mock runtime
Use cases
- 01Deploying and verifying an agent runtime
- 02Managing agent costs and policies
- 03Testing with deterministic offline mocks
MCP support
Server OS includes an MCP server, and its verification checks cover MCP alongside the API, cost controls, governance, workflows, SDK, and CLI.
Configuration
Local server config is the same for every client. Copy it into your client’s MCP config file.
We haven’t collected a package or remote endpoint for this server. Follow the instructions in the project README.
Tools
Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.
Details
From the project READMEServer OS
Autonomous Agentic Operating System — agents as first-class processes with cost control, governance, observability, MCP, SDK, CLI, and skills. Fail-closed. Verified end-to-end.
A senior engineer who has never seen this repository can, using only the source code and this README.md:
- Deploy the entire system (single command)
- Exercise every major feature
- Verify end-to-end correctness via automated checks
No prior context or tribal knowledge required.
Support Agentic OS Kernels ($99) · Agentic OS Cycle ($0.75) · Public Goods Support
Non-custodial USDC (preferred for agents)
| Network | Address | Explorer |
|---|---|---|
| Base | 0xD3d0E9eDAe3Ac7bb199a8EAA761BdA423b878438 | basescan |
| Ethereum | 0xD3d0E9eDAe3Ac7bb199a8EAA761BdA423b878438 | etherscan |
| Solana | ETQwWf19axArsY493UfC6bxe2BmEzmzvCb58PPnC38A | solscan |
Related: rui · LRSI · openmesha · OpenGOS · agenticarb · x402-cloudflare-starter
🚀 Web Control Plane (Hero Demo — Zero Install)
Canonical file: web/server-os.html (~82 KB)
Open locally:
python -m http.server 8088 --directory web
# http://127.0.0.1:8088/server-os.html
Or Chrome → File → Open File… → web/server-os.html.
Acceptance: open the file → OS tab → Golden evals → run → 12/12 passed.
Offline mock provider, policy, budgets, graph, terminal, and 12 golden evals. The HTML engine is DOM-free below the UI so the eval suite exercises the same path as interactive use.
Quick Start (Production Runtime)
docker compose up --build
# another terminal
bash scripts/verify.sh
- API: http://localhost:8080
- Dashboard: http://localhost:8080/dashboard
- OpenAPI: http://localhost:8080/docs
Default mode is offline mock LLM (deterministic, free). Set SERVER_OS_LLM_MODE=openai + OPENAI_API_KEY for real models.
Surfaces
| Surface | Entry |
|---|---|
| Web control plane | web/server-os.html (offline, self-verifying) |
| REST API | python -c "from server_os.main import run; run()" |
| CLI | server-os status / server-os agents ... |
| MCP Server | server-os-mcp |
| SDK | from server_os.sdk import ServerOSClient |
| Multi-agent workflows | skills/multi-agent-workflow/ |
| Skills | skills/*/SKILL.md |
| AGENTS.md | Coding-agent contract at repo root |
| CI | .github/workflows/ci.yml |
Verify contract
bash scripts/verify.sh
Covers API, cost, governance, multi-agent workflows, SDK, CLI, MCP, skills, and AGENTS.md. 16 checks. All must pass.
Web parity check (manual, ~30s): open full web/server-os.html → run Golden evals → 12/12.
Design principles
- Least privilege by construction (capabilities from intent)
- Cost is a first-class resource
- Observable by default
- Fail closed
- Deployable with zero tribal knowledge
License
Apache-2.0
Comments
New to MCP?
From what the protocol is to installing your first server, these answers cover most first questions.
An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.