VibePod CLI 0.20: Credential profiles, Qwen Code, Freebuff, and local image fallback

2026-08-18 · Release · CLI · Agents

Credential profiles: one agent, many logins

Profiles keep multiple credential sets per agent — for example a Claude subscription login, a separate API-key setup, and an environment prepared for a local Ollama — and switch between them at run time. A profile only swaps the credential directories mounted into the agent container; skills, allowed directories, proxy, and logging stay shared.

vp profile create work
vp run claude --profile work
vp task create claude "summarize the diff" --profile work
vp profile list

Existing credentials in ~/.config/vibepod/agents/ are the built-in default profile — nothing moves when you start using profiles. Without the flag, the profile resolves from VP_PROFILE, then the profile: key in the merged config, then default, so a project can pin its own profile in .vibepod/config.yaml and pair it with per-project env vars. vp doctor <agent> --profile <name> diagnoses a specific profile, and vp profile remove deletes one after confirmation.

Qwen Code joins the agent matrix

Qwen Code is Alibaba's open-source terminal coding agent. It is Claude-Code-inspired and supports OpenAI-, Anthropic- and Gemini-compatible endpoints, Qwen models, and any custom provider. Start it with the full name, the qwen-cli alias, or the new vp q shortcut:

vp run qwen
vp q
# then inside the session: /auth

Settings, provider configuration, personal skills, and session history persist under ~/.config/vibepod/agents/qwen/ and survive container restarts. Authentication runs through Qwen Code's own /auth command — connect an Alibaba Cloud Coding Plan, a third-party API key, or a custom endpoint.

Freebuff joins the agent matrix

Freebuff is a free AI coding agent built on the Codebuff platform. Start it with the full name or the vp fb shortcut:

vp run freebuff
vp fb
vp run freebuff login

Credentials, settings, and chat history persist under ~/.config/vibepod/agents/freebuff/. Freebuff authenticates through its own interactive login flow on first run.

Local image fallback

When pulling the latest agent image fails — registry hiccup, rate limit, or no network — vp run and vp task now warn and fall back to the image already present locally instead of aborting. An explicit --pull still fails hard, so you can force a fresh image when you need one.

Release details