Trust layer for AI agents with deterministic gating, isolated execution, and proof for every action
An AI agent with its own cloud and wallet
Stage
Not given
Not given
Founded
Not given
Not given
Based in
Not given
US
Pricing model
Not given
Not given
Pricing
Not given
Free: $0 forever with your own API keys. Pro: $20 per month with $20 of hosted credit a month (or $240 at once on yearly, save up to 17%). Team: $18 per seat, monthly, 2 to 5 people, with $18 of shared credit per seat monthly. Compute costs about $6 per GB a month from credit, charged by the second.
Key features
Deterministic gate that decides before running language models
Isolated code execution with sandboxing and micro-VMs
Append-only audit logs with SHA-256 manifests and hash-chained verification
Honesty measurement under load across multiple nodes
Fleet management with trust plane for machines
OpenAI-compatible endpoints and MCP support
49% UI grounding capability on 600 tasks
95.2% tool calling accuracy
31.6 milliseconds p50 task latency
0.43% false claim rate on 12,344 recorded actions
Real files, real commands, web access, and browser beside the chat
Memory written to disk as plain text you can edit and delete
Auto mode with safe judgment on your own model
Delegate work to helper agents sharing a task list
Chat apps: Telegram, Discord and Slack integration
Support for MCP servers, plugins and IDE integration
Cloud tasks that keep running after you close the app
Services that stay online with scheduler, webhook and watcher
Web apps with their own live link at letronna.si
What makes it different
Complete auditability with cryptographic proof for every action
No vendor lock-in with open standards and open source
Measured honesty benchmarks with 0.43% false claim rate
Production-ready with verified open source repositories and green CI
Runs locally with your own keys; nothing leaves your machine unless you connect a hosted model
Starts in 7.65 ms with 84% on Harness-Bench
Plain text memory files you can edit, not a black box
Works with any model: OpenAI-compatible, Anthropic, Ollama, llama.cpp or hosted
Pay only for hosted models and cloud compute, not per seat
gamAI
Letronna®