What model choice needs around it

Choosing another model should let you continue useful work on the same source. Making that reliable requires more than adding names to a menu. Coding agents differ in their tools, reasoning controls, authentication, and conversation state, and those differences affect which choices work.

Separate the agent from the model

The agent is the program that plans work, calls tools, and manages a conversation. The model supplies its AI responses. Two agents using the same model can expose different controls and preserve different session information.

Rigless hosted projects offer Rigless, Codex, opencode, Pi, and Oh My Pi. Each integration has supported model combinations. The picker needs to reflect those combinations so a visible choice has a supported execution path.

That also determines when a choice can change. In Rigless, choose the agent for a new project and select supported models before or between runs. To use another agent, start a separate project with your source and the context it needs.

Show controls the combination can honor

Reasoning settings differ between models. One may expose named effort levels, another a thinking budget, and another no user-adjustable control. Authentication and context reporting differ too.

A useful interface translates the supported controls and leaves unavailable information clear. Missing provider quota data should not become a guessed allowance. When a request fails, the error should help distinguish provider authentication, usage limits, and an unsupported selection.

The agent and model guide explains the current controls without promising every agent can use every provider.

Keep provider costs understandable

Included AI usage and a customer's provider account are different sources of model capacity. Rigless includes a limited allowance on each plan. Supported provider keys let you use your own model account, with its charges and rate limits.

Changing that source of AI usage does not remove workspace, storage, or publishing limits. Plans and provider keys explains which limit applies and what to do when you reach it.

Keep the project useful after a switch

Source files, dependency manifests, setup instructions, and repository history provide continuity when the agent changes. Native conversation history and tool state are not a common interchange format across agents.

Keep important decisions in the project and retrieve the source through GitHub or a download. A running service may also require separate data, secrets, and external-service configuration to move.

Rigless's vendor-independence promise rests on those practical choices: select a supported agent, use a supported model provider, and retain useful source. Exercise the export path while the project is small, so continuing elsewhere is a workflow you understand.