v0.1.0 — first alpha

Darwin Agentic Cloud — verifiable compute for AI agents.

Install

pip install darwin-agentic-cloud

Live demo

https://darwin-agentic-cloud.fly.dev/docs

What it does

Every code execution returns a cryptographically signed attestation that binds:

• The exact code that ran (workload hash)
• The output (stdout/stderr/exit code hash)
• The substrate it ran on (substrate identity)
• The cost it consumed (signed cost)
• The signer's identity (Ed25519 public key)

Agents can verify execution receipts cryptographically. Auditors can trust them. Tamper a single byte in the attestation and the signature breaks.

Highlights

• Ed25519 signed attestations with end-to-end tamper-evidence verification
• Docker sandbox with isolation primitives and pre-flight cost cap enforcement
• Three transports over one runtime: CLI (darwin run), HTTP (darwin serve), MCP (darwin mcp serve)
• Persistent SQLite audit trail at ~/.darwin/agenticcloud/attestations.db — queryable via CLI, HTTP, MCP
• Working Claude Desktop integration via darwin mcp install (one command)
• Production Docker image — non-root, healthcheck, /data volume mount
• Public hosted demo with branded Material 3 docs at darwin-agentic-cloud.fly.dev

Distribution

• PyPI: pip install darwin-agentic-cloud — https://pypi.org/project/darwin-agentic-cloud/0.1.0/
• Demo: https://darwin-agentic-cloud.fly.dev/docs
• License: Apache 2.0
• Status: Alpha — API may change before v1.0

Tests

37 passing across smoke / attestation / cost / storage / tamper / integration suites. CI green on Python 3.11 and 3.12.