Compare agents

An owner without developer skills can export the agent and restore it working on a runtime or host the provider does not operate, within minutes of asking and without provider approval. Running the runtime on the owner's own machine counts as not provider-operated.

Claude Code documents no whole-agent export or matching import. /export writes only the current conversation as plain text. Agent state is spread across ~/.claude, ~/.claude.json and project .claude/ files, and no procedure is documented for moving it to another machine. Sources

Export is immediate (the agent is one .adf file; copying it and using Open is documented), but restoring a working agent on another machine takes several documented steps: import the owner seed phrase to unlock sealed keys, re-add a model provider key (app-level keys are not in the file), re-register app-level MCP servers, then review and accept the file. One shortfall (restore is not a single action). Sources

The export contains everything that defines the agent: configuration, instructions, memory, action history, logic, schedules and triggers, files, any identity, and credentials or a way to re-link each integration without any provider-registered client or broker. Derived data (embeddings, indexes, summaries) is accompanied by the source it was derived from.

The only export is conversation text. Several classes of state are held by the provider on claude.ai and are not in any export: claude.ai connectors, synced skills and plugins, and cloud sessions/routines. Anthropic-hosted connectors (e.g. Gmail, Google Calendar, Microsoft 365) can only be linked through the redirect URL registered by claude.ai. Sources

The spec places config, instructions, memory, loops, timers, triggers, lambdas, files, messages, audit, logs, sealed identity keys and per-agent credentials in the .adf file; the project holds nothing. Items outside the file (app-level provider keys, app-scoped MCP credentials, WhatsApp session directory, runtime settings) can be re-linked by the owner without any ADF-registered client (MCP OAuth uses dynamic client registration; subscriptions use OpenAI/xAI client ids). A customised global base prompt is an app setting and does not travel; recorded, not counted, because the default install uses the shipped prompt. Sources

With a non-provider model configured per P4, the exported agent runs indefinitely without any service, account, licence or payment controlled by the original provider, and the owner holds a perpetual, irrevocable right to run it. No provider action, inaction or disappearance can disable or degrade it.

The runtime is a closed binary. Its right to use comes from Anthropic's Consumer Terms, under which Anthropic 'may suspend or terminate your access to the Services ... at any time without notice' and may revise the terms at its discretion. No perpetual or irrevocable right to run is granted. Anthropic also 'doesn't support routing Claude Code to non-Claude models through any gateway'. Sources

The runtime is MIT-licensed with no licence check, account or ADF-run relay; the only ADF-project endpoints (GitHub update check, MCP registry, skill catalog, online docs) are optional, have off switches or bundled fallbacks, and are not needed to start or act. Mesh uses LAN, tailnet or owner-chosen addresses. Desk grade; a provider-blocked run is required at certification. Sources

The owner can choose any model that meets a documented minimum, including a local open-weight model, without provider involvement, and every capability of the agent (tools, triggers, messaging, memory, automation) keeps working. A swap may lower reasoning quality but never removes a capability.

ANTHROPIC_BASE_URL, gateway model discovery and CLAUDE_CODE_MAX_CONTEXT_TOKENS are documented and let an Anthropic-format endpoint serve other model IDs. However, Anthropic states it doesn't support routing Claude Code to non-Claude models, so local open-weight models work only through an unofficial workaround. Features tied to Anthropic API betas degrade on non-first-party hosts, e.g. MCP tool search is disabled by default. Sources

Tool calling is the documented minimum and local open-weight models (Ollama, LM Studio, vLLM, llama.cpp) are first-class provider options; triggers, timers, lambdas and middleware run in the runtime. However, video input is implemented by injecting OpenAI-format video_url parts and is documented only for hosted providers (OpenRouter, Gemini), and no test against a local model is published, so under G12 at least one capability is not shown to survive a local model. Sources

Whatever the P2 export contains stays obtainable, current to the moment of termination, after account suspension or ban, a lapsed subscription, a price change, an acquisition, or the provider shutting down.

In the default local CLI, configuration, CLAUDE.md, auto memory, skills and transcripts are plain files on the owner's machine, so they stay available after account termination. Skills synced from claude.ai stay on disk once downloaded. Provider-held classes and 30-day transcript pruning are graded in P2 and A1. Sources

State is local-first: the agent is a SQLite file on the owner's disk with no ADF account or hosted copy, so no termination, ban or shutdown by the project can remove access to it. Sources