Run agents on cloud computers and call language and media models
Stage
Not given
Not given
Founded
Not given
Not given
Based in
US
Not given
Pricing model
Open source
Not given
Pricing
Not given
Free in test mode with 2 hours of computer time per day. Live mode: computers $0.0000556 per second ($0.20 per hour), billed by the second while awake. Claude Sonnet 5.5 (default): $1.92 per million input tokens, $9.60 per million output tokens. Language models from $0.096 to $9.60 per million input
Key features
Harpe: chooses what an AI agent reads under a token budget with consistent results
Kibisis: keeps agent memory on your hardware with AES-256-GCM encryption
Argus: append-only record that detects edits with cryptographic verification
No network code by default for local and disconnected use
MIT License with published code, benchmarks and raw results
Run agents from Claude Code, Codex, pi or Ugoko's own agent
Language models from Claude, Kimi, GLM, Grok and GPT
Image, video, audio and speech generation
Per-end-user budgets and metering
Computer sleeps free after 5 idle minutes
Command-line interface with one-command setup
Test mode free with real computer time
No failed generation charges
Webhooks and event streaming
What makes it different
Measured empirical validation across 6,111 questions outperforming full-context models
Zero plaintext on disk with encrypted vaults matching plaintext accuracy
78.3% fewer tokens sent to the model with a 1,200-token budget per turn
Preserves conversational chronology and relational qualifiers under token budgets
Pay only for computer time while awake, with sleep cost-free
One project balance for all: computers, all language models and all media models
Includes Claude Code, Codex and preview harnesses with no separate gateway to build
Failed jobs are free; pay only for successful generations
Perseus Computing
Ugoko