Who it is forA developer who wants coding agent help on private code without sending it to a cloud API.
The problemCoding agents default to cloud models, which cost money per token and send your code off the machine.
The outcomeClaude Code, Codex CLI or OpenCode talking to a model served by Strata on your machine.
What you need first
- Windows, Linux or macOS on supported hardware (NVIDIA RTX 20 to 50, recent AMD Radeon, Intel Arc on Linux, Apple Silicon M1 or newer)
- Plenty of free disk space; DeepSeekV4.1 sets need about 245 to 280 GB
- The model packs download on first run
Steps
- Install Strata and start a model with
strata launch - Run
strata connectto see the setup instructions for your tool - Apply the settings for Claude Code, Codex CLI, OpenCode or another listed tool
- Pick the Coder pack from the Models docs if you mostly write code
- Run a small coding task and compare speed and quality with your cloud model
Prompt for your agent
Using Niko1221/Strata, start a local model, run strata connect, and configure <Claude Code or Codex CLI> to use it following the printed instructions. Then run a small refactor on <file> and report tokens per second.
Caveat. Local models are slower than cloud APIs on most consumer GPUs; the README lists 4 to 25 tokens per second for DeepSeekV4.1 on single cards.
Use this in your AI agent
Ask your agent
Using the Deduped MCP, get the setup steps for Niko1221/Strata and "Point Claude Code or Codex at a local model with strata connect"
Add Deduped to Claude Code once
claude mcp add --transport http deduped https://dedupedai.com/mcp
Every Monday: the 5 hottest repos and exactly how to use them. Free, one email a week, unsubscribe anytime.
Get them freeShare: X LinkedIn Hacker News Email Copy link
Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer