Skip to main content
Register RunInfra chat models in Continue’s local YAML configuration. Support level: preview.
Continue requires explicit —config for safe first-run/rewrite handling. CLI tool execution is not sandboxed; IDE activation, routing, secrets and restoration after native saves need separate verification.on requires --preview. Preview is not a dry run: it writes settings and sends a request using credits. It does not certify those remaining checks.

​
Connect Continue

Have Continue CLI installed, a workspace API key, and available RunInfra credits. The installed CLI contract is Continue 1.5.47, whose command is cn.
The bare agent command is an alias for on. In an interactive terminal, Connect starts login if needed. Review and approve the proposed changes. Connect saves a snapshot, writes the YAML model entries, and makes a real Chat Completions request that uses credits. To choose a model during setup:
See Connect for status, diagnostics, and key replacement.

​
What Connect writes

An absolute CONTINUE_GLOBAL_DIR replaces .continue directly. Connect writes config.yaml inside it and refuses a relative override. Below, continueHome means that directory, or your home .continue directory when the variable is unset. A new file gets name: RunInfra, version: "1.0.0", and schema: v1. Existing files need readable string name and version fields. An existing schema is optional, but must be v1 when present. Each catalog model gets a runinfra/<slug> name, provider: openai, its bare model id, apiBase: https://api.runinfra.ai/v1, and roles: [chat, edit, apply]. Context and output limits go under defaultCompletionOptions.contextLength and defaultCompletionOptions.maxTokens. Continue CLI 1.5.47 is verified to reject an output limit equal to the context window. Connect caps maxTokens at the smaller of the catalog output limit and half the context window, rounded down, leaving room for input. Matching RunInfra model blocks are refreshed from the catalog, replacing non-catalog options inside those blocks. Other model items are kept byte for byte. Comments inside refreshed blocks are retained as standalone comments. Registering a model does not prove the CLI or IDE selected it. Existing RunInfra models absent from the current catalog remain unchanged. Review them before selecting a model.

​
Key placement

Literal mode writes the key in each managed model’s apiKey. For a secret reference:
Helper mode writes the exact expression ${{ secrets.RUNINFRA_API_KEY }}. Set RUNINFRA_API_KEY securely in the environment that launches the CLI. CLI secret resolution checks the process environment first, then continueHome/.env, the current directory’s .continue/.env, and the current directory’s .env. Startup can already load the current directory’s .env into the process environment. If the secret is still unresolved, the CLI attempts remote secret synchronization at api.continue.dev; it only warns on failure, which does not prove a usable key. Connect does not edit dotenv files or secret stores. Verify IDE secret resolution separately. Restart GUI apps and background services with the required environment. On Windows, set RUNINFRA_API_KEY in the user environment and fully restart the app. GUI and Start menu launches do not inherit a terminal-only assignment.

​
Verify

Continue CLI 1.5.47 completed a verified round trip through RunInfra using the Connect-written config.yaml. The explicit-config onboarding requirement and output-limit rejection are also verified in that CLI version. These observations do not establish IDE activation or complete the remaining preview checks. This manual verification template is docs-only; Connect does not execute it. Use the actual shell-quoted file path in Connect’s launch command. These examples show a home directory named Test User; use your printed path, including any CONTINUE_GLOBAL_DIR override:
Keep --config to bypass onboarding and the headless rewrite described below. Select the intended runinfra/<slug> model and confirm the request destination separately. --config selects a file, not necessarily that model.
--readonly selects plan mode, not a sandbox. Bash, Fetch, and MCP tools remain allowed, and the CLI still sends tool definitions. The endpoint must accept a tools array even for a chat-only catalog model. The no-tools prompt is not enforcement.Use an authorized isolated environment and inspect startup tools and configuration first. An empty scratch directory reduces project-file and dotenv exposure, but does not confine filesystem or network access. Connect’s API probe does not verify Continue’s tools, permissions, or destination.

​
Restart

Connect prints the written file’s real path, shell-quoted, in both the launch command and restart hint. For the example POSIX path above:
For the example Windows path:
Keep --config until continueHome/.onboarding_complete exists and afterward to avoid headless rewrites. Select the intended runinfra/<slug> model and confirm its request destination. Verify IDE local configuration and model selection separately after reload, then start a fresh chat.

​
Troubleshooting

The CLI ignores capabilities, useResponsesApi, and useLegacyCompletionsEndpoint, so Connect does not write them. Their omission is not proof of IDE routing or tool behavior.

​
Revert

With intact state and snapshots, off restores the original bytes from before the first on in the active connection. If config.yaml did not exist then, Connect removes it. Later on calls and key rotations keep that restore point. If Continue or another editor rewrote the file, Connect refuses to overwrite it and retains the origin snapshot. Back up and review the edits before choosing to discard them:
Connect saves the current contents in a recovery snapshot and prints its id before restoring the origin. Keep the snapshots, since CLI and IDE saves can change comments and layout. If the origin snapshot is missing, cleanup does not guarantee original-byte restoration; see Connect recovery. Restore manual environment or secret-store changes separately. File restoration does not create, remove, or restore the onboarding marker.

Connect

Set up, inspect, and restore your coding agent.

Chat completions

Read the request contract.

Models

Find model ids and context windows.