OpenCode Compat Shim
ArcticWinterSturm/opencode-compat-shim
OpenAI-style streaming proxy that lets Hermes Agent use OpenCode's free-tier models
OpenCode Compat Shim is a local OpenAI-shaped SSE proxy that routes OpenCode's free-tier models through OpenCode's own CLI so Hermes Agent can call them. It gives Hermes Agent users a way to try those free models as a custom provider.
What OpenCode Compat Shim does
This is an OpenAI-shaped SSE proxy that tunnels OpenCode's free-tier models through OpenCode's own CLI so Hermes Agent and other OpenAI-compatible clients can use them. The README lists big-pickle, muse-spark-1.3-contributor-free, mimo-v2.5-free and deepseek-v4-flash-free. It exists because OpenCode ties the free tier to desktop-app identity, so plain HTTP requests are rejected with a 403, and only the signed CLI binary is authenticated.
Version 2 fixes three problems in the first version: it now streams SSE frames instead of returning one JSON blob, keeps conversation history across turns, and fixes a 404 on the default client path. The proxy runs on port 18788 with a /health endpoint, started from PowerShell in the README. To use it with Hermes you add a custom provider to config.yaml with base_url http://127.0.0.1:18788/v1, enable streaming for that provider, and arrange for the proxy to start automatically. The README says these three Hermes-side steps remain.
Key features
- OpenAI-style /v1/chat/completions endpoint with SSE streaming
- Conversation history kept across turns
- Health check endpoint on port 18788
- Four OpenCode free-tier model names supported
- Custom provider configuration for Hermes config.yaml
When to use it
- Try OpenCode free-tier models from inside Hermes Agent
- Point any OpenAI-compatible client at the same local proxy
- Test the proxy over raw HTTP with curl before wiring Hermes
Who it is for: Hermes Agent users on Windows who already have OpenCode installed and want to use its free models in Hermes.
How it fits with Hermes Agent
Built for Hermes Agent as a custom OpenAI-compatible provider endpoint, and usable with other OpenAI-compatible clients.
Requirements: The OpenCode desktop app installed with its CLI authenticated, and aiohttp in the Python that runs the proxy.
Note: The README says the proxy is verified over raw HTTP but that three Hermes-side steps are still needed for end-to-end use, and it depends on the OpenCode desktop app's authenticated CLI.
FAQ
What is OpenCode Compat Shim?
OpenCode Compat Shim is a local proxy that translates OpenCode's CLI protocol into an OpenAI-style streaming API, so Hermes Agent can use OpenCode's free-tier models.
Does OpenCode Compat Shim work with Hermes Agent?
It is built for Hermes Agent, but the README says three Hermes-side steps remain for end-to-end use: registering the custom provider, enabling streaming and auto-starting the proxy.
What do I need to run OpenCode Compat Shim?
You need the OpenCode desktop app installed with its CLI authenticated, and aiohttp installed in the Python that runs the shim.
Similar models for Hermes Agent
All modelsPer-token cost accounting for local LLM inference rigs in Hermes Agent
BlockRunAI ClawRouter for HermesPlugin that adds BlockRun's ClawRouter LLM router to Hermes as one provider, paid per request
profbernardoj Morpheus SkillDecentralized AI inference skill with 30+ models via Morpheus, for OpenClaw, Hermes Agent and others
moreoronce Hermes ZCode GLM PatchCompatibility patch that fixes misleading 429 code 1305 errors when Hermes Agent uses GLM-5.2 on Z.AI
zeyxx hermes-antigravityGoogle Antigravity and Cloud Code Assist inference provider plugin for Hermes Agent using OAuth sign-in
pooops15 muse-bridgeBridge that exposes the Muse agent as an OpenAI-compatible ms/muse model inside 9Router for Hermes
Related guides: How to install Hermes Agent