# Claude in Scient Claude Code is Anthropic's agent for working with project files, code, commands, and connected tools. Scient can use an existing Claude installation or manage a private copy, then connect it to a Claude or Anthropic Console account. Installing the tool and signing in are separate steps. For the behavior shared by all assisted providers, see [Providers in Scient](./providers.md). ## The Easiest Setup Path When Claude is not ready, choose it from **Choose your AI** in the composer or open **Settings > Providers**. The same guided surface appears in both places: 1. Scient preserves and verifies a healthy custom or system Claude installation when one exists; 2. otherwise, the local desktop app can install a qualified official Claude Code executable in Scient's private app data; 3. the available sign-in button starts Claude Code's official account flow and opens its secure page in your browser; and 4. Scient asks Claude for fresh account and model state before reporting it as ready. Connecting another AI does not hide Claude. Every enabled provider remains available in the model picker, and you can choose any Claude model reported for the connected account. **Sign in to Claude** uses Claude Code's official sign-in flow for an existing Claude.ai Pro, Max, Team, or Enterprise subscription. **Use Anthropic Console** remains available as an alternative for an Anthropic Console account. Existing API-key, Bedrock, Vertex, Foundry, and custom-endpoint setups remain advanced provider configurations rather than being rewritten by the guided account flow. Most browser sign-ins return to Claude automatically. If the browser itself displays a one-time code, expand **Browser showed a code?** and paste it into the compact recovery field. This is an occasional fallback for environments where the local callback cannot complete, not a normal extra step. The code is sent only to the live Claude process and is not saved as provider state. Scient never asks for or receives your Claude password. Claude Code continues to own its credentials, refresh, expiry, and revocation. Claude's local account status can occasionally outlive a remotely revoked token. If a live Claude request reports that exact condition, Scient stops reusing the failed session and returns the provider to **Sign in required**. Sign in again from the same guided surface. Scient does not send a test prompt or consume model quota merely to validate the account after login. ## Computers And Installations Scient never silently replaces a healthy custom or system installation. A Scient-managed copy is private and removable, and removing it does not remove a system Claude install, a custom binary, or Claude-owned account data. For a Scient-managed copy, expand Claude in **Settings > Providers** to repair or remove the private runtime. If the executable stops working, **Manage** opens directly on that recovery path instead of asking you to sign in again. Managed installation is offered only when the local desktop server reports a qualified artifact for its exact operating system and architecture. Remote clients continue using the runtime administered on their server. The app displays the qualified version and download details before installation. When a newer qualified stable release is available, **Update** preserves the current working copy until the replacement passes local verification and activation. **Repair** installs the latest qualified release (or the installed one, when that is newer), so it can also bring a newer release. ## Advanced And Multiple-Account Setups The remaining sections are for people who deliberately want more than one Claude setup or an external Claude-compatible route. For Codex, see [Codex](./providers-codex.md). Common reasons: - use separate work and personal Claude accounts - try a different Claude Code configuration without disturbing your main setup - run Claude through a router such as Claude Code Router - use external providers exposed through a Claude-compatible workflow ## I Only Use One Claude Account Use the default provider. The guided Scient flow is recommended. The equivalent Claude subscription recovery command is: ```bash claude auth login --claudeai ``` Use `claude auth login --console` instead when you deliberately want an Anthropic Console account. In Scient Settings, your Claude provider can stay like this: ```text Display name: Claude Binary path: claude CLAUDE_CONFIG_DIR path: empty ``` An empty `CLAUDE_CONFIG_DIR path` means Scient uses Claude Code's normal config directory. When you set this field, Scient points Claude Code at that directory with the `CLAUDE_CONFIG_DIR` environment variable. It does not change `HOME`, so your system keychain and the rest of your environment stay as they are. Claude Code's verbose mode can stay enabled when you use Claude for text generation, including thread titles, branch names, commit messages, and pull request descriptions. On a remote connection, Scient uses the Claude configuration on the connected server. ## Connected tools after upgrading Scient's Claude tool connection is now named `scient` rather than `t3-code`. Claude's saved “always allow” choices for the old connection do not carry over, so Claude may ask you to approve the connected tools again. Review those requests and approve the tools you want to use. Your conversation history and selected permission mode are kept. ## Compact long conversations In Settings, open your Claude provider and set **Auto-compact after** to a token count between `100000` and `1000000`. For example, `300000` compacts the conversation into a summary once it reaches about 300,000 tokens, without changing the model's context window. Leave the field empty to keep Claude Code's default behavior. You can also send `/compact` in an existing conversation. Claude Code also compacts automatically. To see context usage, see [commands and skills](./composer.md#commands-and-skills). When you return to a large thread after Claude's prompt cache expires, a **Compact** chip with the thread's token count shows next to the send button. While it is on, Enter summarizes the history first, then sends your message. Click the chip to switch it to **Keep** and keep the full history for that message. The mobile app shows no chip: an eligible message sent from it summarizes first. The cache lasts about an hour on a Claude subscription and 5 minutes with an API key, Bedrock, or Vertex. See [commands and skills](./composer.md#commands-and-skills) for using composer commands. Claude can also show its own resume prompt when you continue an old session. **Compact idle threads** in **Settings → Providers → Advanced** controls this for threads over 200,000 tokens. Each Claude provider's settings can override it: - **When you return** shows the **Compact** chip described above. This is the default. - **Before the cache expires (beta)** tries to compact an idle thread shortly before its prompt cache expires, while the history is still cached. The compaction appears in the thread as a `/compact` turn sent by Scient. If the window is missed, such as when you enable this mode after the cache has expired or the server restarts before compaction runs, it falls back to **When you return**. Switch the chip to **Keep** to keep the full history for that message. - **Never** skips idle compaction and Claude's resume prompt. **Auto-compact after** still applies, and you can still compact by hand. ## Usage limits If Claude reaches a subscription usage limit mid-turn, the conversation shows the limit and the remaining wait when Claude reports a reset time. Claude may keep the turn open until the window reopens. Wait for that reset, or stop the turn and continue later; the warning timestamp indicates when its displayed wait began. ## Where Claude Skills Are Loaded Scient looks for local Claude skills in the Claude config directory's `skills` folder and `/.claude/skills`. If the same skill name exists in more than one folder, the one in the Claude config directory wins, the same way Claude Code resolves it. Use `$` in the composer to select a skill. A skill set to `off` in Claude Code's `skillOverrides` is left out of both composer menus. Enabled installed Claude plugins also contribute skills under names such as `plugin:review`. Skills marked `disable-model-invocation` can still be started by you. Invoke those one per message: Claude directly runs only the last named skill and may try to start earlier ones through its Skill tool, which refuses skills reserved for manual invocation. ## Task lists To show Claude task lists in the web and desktop apps, add `CLAUDE_CODE_ENABLE_TODO_TOOLS=1` under **Variables** in the Claude instance's **Environment** section in **Settings > Providers**. Each update shows as a Tasks row in the thread; open it to see every step. Unfinished tasks also stay in a bar above the composer until they are done, the session changes, or you dismiss them. ## I Want Work And Personal Claude Accounts Use a different Claude config directory for each account. Example: ```text default config dir work account ~/.claude_personal_home personal account ``` ### Set Up The First Account Log in with the first Claude subscription: ```bash claude auth login --claudeai ``` In Scient Settings: ```text Display name: Claude Work Binary path: claude CLAUDE_CONFIG_DIR path: empty ``` ### Set Up The Second Account Log in with a separate config directory: ```bash mkdir -p ~/.claude_personal_home CLAUDE_CONFIG_DIR=~/.claude_personal_home claude auth login --claudeai ``` Use `--console` instead of `--claudeai` when that provider should use an Anthropic Console account. Use `CLAUDE_CONFIG_DIR`, not `HOME`. Setting `HOME` writes the login to `~/.claude_personal_home/.claude`, which is not where Scient looks. Then add another Claude provider in Scient: ```text Display name: Claude Personal Binary path: claude CLAUDE_CONFIG_DIR path: ~/.claude_personal_home ``` On mobile, tap **Add Claude account** in **Settings > Providers** and enter a name and the config folder. The login itself still happens on the environment's machine. Use the email shown in Settings to confirm each provider is using the intended account. The email is visible initially so accounts can be distinguished; select it to hide or show it. ## Can I Switch Claude Accounts In An Existing Thread? Usually, no. Scient only offers Claude providers that use the same config directory for an existing thread. A different config directory is treated as a different Claude environment. This is different from the recommended Codex setup. Claude Code keeps account and local state across multiple files under its config directory, so Scient keeps separate config directories isolated instead of trying to share part of the state. ## I Want To Use OpenRouter Use this when you want Claude Code to talk to OpenRouter directly, without running a local router. This is the simplest external-provider setup. OpenRouter provides a Claude Code integration through Claude's Anthropic-compatible environment variables. ### Configure A Claude OpenRouter Provider Add or edit a Claude provider in Scient Settings: ```text Display name: Claude OpenRouter Binary path: claude CLAUDE_CONFIG_DIR path: ~/.claude_openrouter_home ``` In that provider's Environment variables section, add: ```text ANTHROPIC_BASE_URL https://openrouter.ai/api ANTHROPIC_AUTH_TOKEN sk-or-... Sensitive ANTHROPIC_API_KEY Empty value ``` Mark `ANTHROPIC_AUTH_TOKEN` as sensitive. Scient stores the value as a server secret and does not send it back to the app after saving. If you want this setup isolated from your normal Claude account, create that home first: ```bash mkdir -p ~/.claude_openrouter_home ``` If you previously used the same Claude home with a normal Anthropic login, run `/logout` in a Claude Code session for that home before using OpenRouter. Otherwise Claude Code may keep using cached Anthropic credentials instead of the OpenRouter token. ### Pick OpenRouter Models Select the model you want in Scient. For an OpenRouter model outside the built-in list, open that Claude provider in **Settings > Providers**, add its full model ID with **Add custom model**, and select it in the chat model picker. Custom models may have fewer effort, thinking, or context controls than built-in models. `ANTHROPIC_DEFAULT_*_MODEL` variables map Claude Code aliases such as `sonnet`; they do not replace an explicit model ID selected in Scient. They remain useful when you want stable defaults for Claude Code's internal model roles. Example: ```text ANTHROPIC_DEFAULT_OPUS_MODEL anthropic/claude-opus-4.6 ANTHROPIC_DEFAULT_SONNET_MODEL anthropic/claude-sonnet-4.6 ANTHROPIC_DEFAULT_HAIKU_MODEL anthropic/claude-haiku-4.5 CLAUDE_CODE_SUBAGENT_MODEL anthropic/claude-sonnet-4.6 ``` Add those to the same provider's Environment variables section if you want stable model choices. ### Verify OpenRouter Is Being Used Open a Claude session and run: ```text /status ``` You should see the Anthropic base URL set to: ```text https://openrouter.ai/api ``` You can also check the OpenRouter activity dashboard for requests from your API key. ### Common OpenRouter Mistakes - Use `https://openrouter.ai/api`, not `https://openrouter.ai/api/v1`, for Claude Code. - Set `ANTHROPIC_AUTH_TOKEN` to your OpenRouter API key. - Set `ANTHROPIC_API_KEY` to an empty string so Claude Code does not try to use an Anthropic login. - Put these variables on the Claude provider instance, not in global shell startup files. OpenRouter's setup can change over time. Use its upstream Claude Code guide for the current details: . ## I Want To Use Claude Code Router Claude Code Router is useful when you want a local routing layer with more control than a direct OpenRouter setup. Scient does not need a special Claude Code Router provider. Treat the router as a Claude environment: give a Claude provider its own `CLAUDE_CONFIG_DIR path`, and put whatever variables the router tells you to export into that provider's Environment variables section. Mark tokens and API keys as sensitive. ```text Display name: Claude Router Binary path: claude CLAUDE_CONFIG_DIR path: ~/.claude_router_home ``` Follow the upstream project's README for the router's own install, startup, and configuration steps: . ## I Want Different Claude Settings, Not A Different Account Create another Claude provider with the same account if you want a named preset. Examples: - "Claude Default" - "Claude Router" - "Claude Experimental" If the preset needs different Claude files, give it a different `CLAUDE_CONFIG_DIR path`. If it needs different API keys, base URLs, or router settings, use Environment variables. Do not put environment variable assignments in `Launch arguments`.