orchesis 0.1.1

AI Agent Control Plane — transparent HTTP proxy for security, cost optimization, and reliability

Orchesis is a transparent HTTP proxy between AI agents and LLM APIs. Every request passes through a 17-phase detection pipeline. Zero dependencies. MIT license. AI Agent -> [Orchesis: 17 phases] -> LLM Provider (OpenAI, Anthropic...)

Installation

# Core (zero dependencies)
pip install orchesis

# With integrations (Slack, Telegram, webhooks)
pip install orchesis[integrations]

orchesis quickstart --preset openclaw

One line change:

# Before:
client = OpenAI(base_url="https://api.openai.com/v1")

# After:
client = OpenAI(base_url="http://localhost:8080/v1")
# ↑ 17 security phases now active

How it works

graph LR
    A[AI Agent<br/>OpenClaw/CrewAI/LangChain] -->|HTTP request| B
    B[Orchesis Proxy<br/>17-phase pipeline<br/>localhost:8080] -->|filtered request| C
    C[LLM Provider<br/>OpenAI/Anthropic/Google]
    B --> D[Dashboard<br/>Metrics & Alerts]

What Orchesis does

	Security	Cost	Reliability	Observability
	17-phase detection. Prompt injection, credential leaks, tool abuse. 25 signatures.	Context compression 80-90%. Semantic cache. Thompson Sampling routing.	Auto-healing. Circuit breakers. Loop detection. 6 recovery actions.	Real-time dashboard. Flow X-Ray. Agent Reliability Score.

By the numbers

Metric	Value
Pipeline phases	17
Threat signatures	25 across 10 categories
Token savings	80-90%
MAST coverage	78.6%
OWASP coverage	80%
Auto-heal actions	6
Tests passing	2,927
Dependencies	0 (stdlib only)

Free MCP Security Scanner

Check your MCP configuration for security issues:

→ orchesis.io/scan

Or via CLI:

orchesis audit-openclaw

---

Contributing

Note: dashboard/dist/ is intentionally committed for zero-setup deployment. Run npm run build in dashboard/ to rebuild from source.

Website · Documentation · MCP Scanner · Blog

MIT License · Built with ❤️ and zero dependencies