aider can use SoAI through its OpenAI-compatible Chat Completions API. aider does not provide a native MCP client, so this connection supplies model inference rather than SoAI tools.
Create a key and choose a model
In the SoAI WebUI, select the Settings button at the right of the header, choose Settings, and open
OpenAI. This tab appears only to an Administrator with API key management
(OPENAI_API_ADMIN). Select Create key and copy the key when it is shown. Then open Models and copy the exact ID
of a model that supports Chat Completions; this requires Model read access
(MODEL_READ). Coding quality depends on the selected model's context, reasoning, and
code-editing capability.
Run aider through SoAI
export OPENAI_API_KEY='YOUR_SOAI_API_KEY' export OPENAI_API_BASE='https://SOAI_HOST:5090/v1' export SSL_CERT_FILE='/path/to/ca-bundle-containing-soai.pem' aider --model openai/YOUR_MODEL_ID
The openai/ prefix tells aider's LiteLLM layer to use the OpenAI-compatible adapter; it is not part of
the model ID. For an unknown model, aider may warn that it lacks pricing or context metadata. Supply an aider
model-settings file (for example, for a custom model) when accurate context limits or edit formats are needed rather than renaming the model.
MCP support and troubleshooting
aider does not accept an MCP server configuration. Use an MCP-capable client when the agent must call SoAI tools; creating a SoAI MCP token does not make tools available to aider.
Troubleshooting
A 401 response means the SoAI API key was not loaded. A 404 generally means the base URL lacks /v1
or the model ID is wrong. A remote LiteLLM model-price download warning is independent of inference. Trust the
server's CA through a combined CA bundle; never use aider's option to disable certificate verification.