Documentation
Quickstart
Install TokForge, download your first model, create a Forge Config, and run your first benchmark in under 5 minutes.
API Reference
120+ endpoints across 11 handler groups for remote control of inference, models, config, benchmarking, navigation, downloads, settings, and AI agent workflows.
ForgeLab & Methodology
Four optimization tiers, AutoForge config sweeps, speculative decoding benchmarks, cross-device comparison, and reproducible methodology.
Speculative Decoding NEW
Up to 2x faster inference with draft-verified acceleration. Auto-detected model pairings, smart device routing, and live speed indicators.
Memory System
Persistent per-character memory with background extraction, hybrid retrieval, knowledge graphs, character impressions, and document import.
Guides 12 NEW
Step-by-step walkthroughs for offline AI setup, character cards, document RAG, AutoForge tuning, API automation, and more.
Privacy Policy
Zero data collection. No analytics. No telemetry. All inference on-device. Read the full policy.
Terms of Service
Usage terms for the TokForge platform and beta program. Licensing under review.