Proof of what an AI agent was allowed to do, and what it then did
Coding agents on cloud machines with a record of every session
Stage
Startup
Not given
Founded
2026
Not given
Based in
Not given
Not given
Pricing model
Open source
Free trial
Pricing
Not given
Free trial: two machines, up to 2 vCPU and 4 GiB. Cuna Inference pricing: models available at varying rates, such as GLM-5.2 at $1.00 per million input tokens and $4.00 per million output tokens.
Key features
Signed permissions with passkey authentication
Receipts for each agent action with signatures
Free checker tool to verify compliance
Quantum-safe signatures (Ed25519 and ML-DSA-87)
Permission narrowing for helper agents
Permission cancellation with signed acknowledgement
JavaScript and Python libraries
LangChain and Pydantic AI middleware
Isolated session execution with bounded authority
Web console for machine management
Complete action recording and visibility
OpenAI-compatible Chat Completions API
OpenCuna terminal agent
Multiple model provider support
Cuna Memory for agent state persistence
Token-based pay-as-you-go pricing
What makes it different
Only open-source project combining machine-checkable limits, permission narrowing for helpers, and signed permission cancellation
Built entirely from published standards with no hand-made components
Free and no dependencies, sends nothing anywhere or holds money
Distinguishes between record integrity and whether agent stayed within permission
All published versions stay available for backward compatibility
Your Agent. Not a Vendor's.
The work ends. The proof remains.
Complete session records and operational transparency