AI Agent Hub
Back to plugins
🤖

dsh-freeroute

Model Inference Updated 2026.08.27

Run the following command in DeepSeek Harness:

dsh plugin install 0xrushmoon/dsh-freeroute

Paste the following prompt into your AI chat to install this plugin:

Run dsh plugin install 0xrushmoon/dsh-freeroute in the DeepSeek Harness terminal to install; full source is available at https://github.com/0xrushmoon/dsh-freeroute

About this plugin

Every major LLM provider offers a small free inference tier, but a single account hits rate limits quickly and a dead key forces you to manually reconfigure or switch upstreams. dsh-freeroute aggregates OpenCode Zen, B.AI, OpenRouter, SenseNova and other free-quota providers behind a single freeroute model entry in dsh. Its core mechanism is multi-key rotation with automatic failover: each upstream accepts up to 8 API keys cycled in order, and if any provider fails before the first token reaches your session, the request transparently moves to the next candidate so the caller never sees a break.

A built-in local OpenAI-compatible endpoint lets any Agent, CLI tool, or browser extension that accepts a custom base URL ride the same free pool with no extra API key and CORS open. The settings panel embeds as a sub-tab inside dsh, offering per-upstream enable/priority/connectivity-test/health cards, one-click integration wizard, remote catalog sync hosted on Cloudflare Pages or R2 for zero-cost model list updates, and a custom gateway form for local uni-api, new-api or LiteLLM setups. A non-standard gateway only needs a base URL and an optional chat-path override.

Ideal for developers who want to try multiple models at zero marginal cost without scripting their own round-robin and recovery logic, and for teams that already run Agent pipelines and want a few free redundant nodes behind their existing tools.

Use Cases

  • Try multiple models at zero marginal cost without scripting round-robin or recovery logic
  • Add free redundant inference nodes to an existing Agent pipeline
  • Drive inference from browser extensions or CLI tools via a local OpenAI-compatible endpoint

Best For

  • Individual developers seeking multi-model access at no extra cost
  • Teams running Agent pipelines that need redundant free inference nodes
  • Users with local gateways (uni-api, new-api, LiteLLM) wanting unified management