Install
$ agentstack add mcp-augustus-so-protocol ✓ scanned · ✓ verified, works with Claude Code, Cursor, and more.
Security review
✓ PassedNo issues found. Passed automated security review. · v0.1.0 How review works →
- ✓ Prompt-injection patterns
- ✓ Secret / credential exfiltration
- ✓ Dangerous shell & filesystem operations
- ✓ Untrusted network calls
- ✓ Known-malicious package signatures
What it can access
- ✓ Network access No
- ✓ Filesystem access No
- ✓ Shell / process execution No
- ✓ Environment & secrets No
- ✓ Dynamic code execution No
From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.
Verified badge
Passed review? Show it. Paste this badge into your README, it links to the public security report.
Reliability & compatibility
Declared compatibility
Compatibility is declared by the source manifest. End-to-end runtime verification is coming, see below.
We're building live execution health for every listing: tool-call success rate, median latency, uptime, and last-checked timestamps, measured, not self-reported. It isn't live yet, so we don't show numbers we can't stand behind.
How agent discovery & health will work →About
The Augustus Protocol
Protocol version 1.2.0
An open standard for driving a business through an AI operating layer from any coding agent.
Point Claude Code, Cursor, or Codex at Augustus. It grounds itself in your business, audits a process, designs an AI system for it, tests that system against your real data, and runs it with a human in the loop. You approve what matters. It does the rest.
This repository is the standard itself: the lifecycle, the gates, and the full tool interface, in plain language and in machine-readable schema. It is open so anyone can read exactly how the protocol works, drive it from any MCP client, or implement a compatible engine of their own.
Why this is open
The way you rebuild a company around AI should be a shared, inspectable standard, not a black box you have to trust. So the standard is here in full: every stage, every gate, every tool, in plain language and in machine-readable schema.
A spec cannot hand you the part that runs it: a hosted engine, the validation that refuses anything unproven, and the reference implementation that keeps it reliable at scale. The standard is open. The engine is Augustus.
The lifecycle
The protocol takes one process from raw to live in eight stages. An agent drives every step. The owner answers questions and approves decisions.
- Foundation. The agent reads your website and everything you hand it, and grounds itself in what your business is.
- Audit. For the process you pick, the agent writes a deep audit on its own model, against a fixed methodology.
- Ground. The protocol refuses to build on a guess. If a fact is missing, it asks you in plain language until the audit stands on real answers.
- Build. The agent designs the automation using only tools that are connected, and submits it into safe preview.
- Test. The system runs against real data in preview. It touches nothing real until it is proven.
- Go live. A proven system goes live on its trigger. Every irreversible action still waits for a human.
- Wire the trigger. The agent connects your real event source so the system fires.
- Operate. Approvals, decisions, and revisions run through the agent. The system improves as it runs.
The full contract is in [SPEC.md](SPEC.md). Every tool is in [TOOLS.md](TOOLS.md), with machine-readable schema in [schema/tools.json](schema/tools.json).
Where the work runs
- Designing runs on your agent. Writing the audit and designing the automation run on your own coding agent. The engine hands it a brief: the methodology, your grounded context, and an output schema. Your agent produces the result. The engine validates it and enforces every gate.
- Running runs on your model. A live system runs later with no agent present, on your own AI key. Augustus is the layer, not the model. You bring the model.
Quickstart
The reference engine runs as an MCP server. Connect your coding agent to it and start.
Claude Code
claude mcp add --transport http augustus https://app.augustus.so/api/mcp
Cursor, Codex, or any MCP client
Add the server to your MCP config:
{
"mcpServers": {
"augustus": {
"url": "https://app.augustus.so/api/mcp"
}
}
}
Then tell your agent: "Set up my business on Augustus." It takes it from there. Anyone can run the audit for free. Building a real system creates a tracked account, which the agent handles in the conversation.
What is in this repo
| File | What it holds | | --- | --- | | [SPEC.md](SPEC.md) | The protocol: principles, the eight-stage lifecycle, and the gates. | | [TOOLS.md](TOOLS.md) | Every tool an agent calls, grouped by stage, with parameters. | | [schema/tools.json](schema/tools.json) | Machine-readable schema for all 45 tools. | | [CONTRIBUTING.md](CONTRIBUTING.md) | How to propose changes to the standard. | | [CHANGELOG.md](CHANGELOG.md) | Version history. |
Reference engine
Augustus is the reference implementation. It hosts the runtime and the validation. How to connect and what it costs live there, not in this standard.
Licensed under [MIT](LICENSE).
Source & license
This open-source MCP server is cataloged on AgentStack and links to its original source — we do not rehost the code.
- Author: augustus-so
- Source: augustus-so/protocol
- License: MIT
Install and usage instructions live in the source repository linked above.
Reviews
No reviews yet, be the first.
Write a review
Versions
- v0.1.0 Imported from the upstream source.