The best way to control Claude Code, Codex and more from any computer or phone
| 1 | # ACP providers |
| 2 | |
| 3 | T3 Code can run installed ACP commands or coding agents from the official |
| 4 | [ACP Registry](https://agentclientprotocol.com/get-started/registry). Registry agents bring their |
| 5 | own models, tools, and sign-in, while T3 Code provides projects, threads, checkpoints, and task |
| 6 | delegation. |
| 7 | |
| 8 | Agents that T3 Code supports as their own providers, such as Codex, Claude, Cursor, Grok, OpenCode, |
| 9 | Pi, and [Antigravity](./providers-antigravity.md), don't appear in the Registry list. Enable them in |
| 10 | **Settings → Providers** instead. |
| 11 | |
| 12 | T3 Code prefers the current ACP v2 preview protocol and uses its richer messages, usage, plans, |
| 13 | configuration, compaction, and agent-terminal updates when the agent supports them. It also |
| 14 | negotiates ACP v1 for Registry agents that have not migrated yet, so they continue to work through |
| 15 | the same generic integration. |
| 16 | |
| 17 | Claude Agent uses the ACP v2 preview by default for Registry and local `claude-agent-acp` |
| 18 | commands. If you encounter preview issues, set `CLAUDE_AGENT_ACP_EXPERIMENTAL_V2=0` in the |
| 19 | provider instance's environment overrides to use ACP v1. |
| 20 | |
| 21 | ## Add an agent |
| 22 | |
| 23 | 1. Open **Settings → Providers**. |
| 24 | 2. Select **Add provider**, then **Show more**. |
| 25 | 3. Search the ACP Registry for the agent and select **Add** on its result. |
| 26 | 4. Confirm the name and instance ID, then complete the agent's sign-in step. |
| 27 | |
| 28 | Search only shows agents that can run on the connected server. Registry agents are third-party |
| 29 | code; review an agent's source and license before adding it. |
| 30 | |
| 31 | ## Add a local command |
| 32 | |
| 33 | In **Settings → Providers → Add provider**, select **Show more**, then **Local ACP command**. Enter the executable |
| 34 | name or path on the selected environment, add one literal argument per row, and choose a display |
| 35 | name. For DeepSeek Harness, use `dsh` with arguments `--profile` and `acp`. |
| 36 | |
| 37 | Local commands require no registry entry or download. Environment overrides apply to the command; |
| 38 | credentials and skills remain managed by the installed agent or its wrapper. Models and reasoning |
| 39 | options come from ACP, including model IDs that contain JSON. T3 Code launches the executable directly |
| 40 | without expanding shell expressions. |
| 41 | |
| 42 | ## Where agents run |
| 43 | |
| 44 | Registry agents always run on the machine that hosts your T3 Code server. That stays true when you |
| 45 | connect through `app.t3.codes`, T3 Connect, or a relay. |
| 46 | |
| 47 | Agents install under `tools/<agent-id>/<version>/` inside T3 home. T3 Code verifies SHA-256 when the Registry entry |
| 48 | provides one; entries without a checksum retain the Registry's HTTPS distribution guarantee. |
| 49 | Registry `npx` and `uvx` packages use T3-owned npm prefixes and Python tool directories at the exact |
| 50 | version published by the Registry. Their commands are available in a new server terminal for |
| 51 | sign-in and direct use. Removing an agent's last provider instance removes T3-managed binary files |
| 52 | but keeps package installs. To use an existing local binary, set **Executable override** explicitly. |
| 53 | |
| 54 | ## Signing in |
| 55 | |
| 56 | Open the agent's account section in **Settings → Providers** on web or desktop. Choose |
| 57 | **Sign in** and, if the agent offers several methods, select one. For an installed, configured |
| 58 | provider, mobile also offers **Settings → Providers**. |
| 59 | |
| 60 | Browser sign-in shows the agent's URL and waits for you to open or copy it before telling the |
| 61 | agent to proceed. The page opens on your device, while the agent runs on the environment. |
| 62 | Terminal methods run in an in-app terminal on that environment. On mobile, send responses |
| 63 | through the terminal response field. T3 Code reconnects after terminal login and waits for the |
| 64 | agent to confirm sign-in before reporting success. You can cancel or retry an expired attempt. |
| 65 | |
| 66 | Browser and terminal sign-in leave credentials in the agent's own store. Agents that use API keys |
| 67 | through environment variables still take those keys in the instance's environment settings. |
| 68 | A configured environment or agent home applies to both sign-in and chat. |
| 69 | |
| 70 | If the agent advertises logout, choose **Sign out**. Changing a shared agent login stops running |
| 71 | threads for instances of that same agent on the environment. Thread history and workspace files |
| 72 | are kept. ACP does not describe account isolation, so adding another instance does not guarantee |
| 73 | a separate account; use the agent's own configuration to isolate accounts when supported. |
| 74 | |
| 75 | If an agent cannot complete its advertised flow remotely, its CLI remains available on the |
| 76 | server. For example, Codex supports `codex login --device-auth`, and Grok Build supports |
| 77 | `grok login --device-auth`. |
| 78 | |
| 79 | ## Models and options |
| 80 | |
| 81 | The model picker lists the models the agent reports. If the agent reports none, or hides some, add |
| 82 | your own model IDs under **Custom models** in the instance settings. |
| 83 | |
| 84 | Other settings the agent offers, such as reasoning effort or approval mode, appear in the |
| 85 | composer's model picker, beside the model list, like they do for built-in providers. Models and options that change while an agent |
| 86 | is running update the picker without waiting for another provider probe. |
| 87 | |
| 88 | ## Commands and skills |
| 89 | |
| 90 | Slash commands the agent provides appear under **Provider** in the `/` menu while a session is |
| 91 | running. Commands the agent names with a `$` prefix appear in T3 Code's `$` skill menu instead. |
| 92 | |
| 93 | ## Native sessions |
| 94 | |
| 95 | When an agent advertises ACP session listing and import support, expand its provider card, |
| 96 | choose a project, and select **List sessions**. Importing one creates a deterministic T3 thread |
| 97 | backed by that native session. Importing the same agent session again returns the existing thread, |
| 98 | including when another client performs the import. |
| 99 | |
| 100 | The thread keeps the native session's title and last-update value as provider metadata without |
| 101 | overwriting a title you set in T3 Code. Agents that report ACP context usage drive the thread details |
| 102 | context window and cumulative cost metadata. See |
| 103 | [commands and skills](./composer.md#commands-and-skills) to show it on web and desktop. |
| 104 | Text resources and links render as assistant output; |
| 105 | binary resources, images, and audio that T3 Code cannot render yet appear as explicit placeholders |
| 106 | instead of disappearing. |
| 107 | |
| 108 | ## Permissions and terminals |
| 109 | |
| 110 | Registry agents read files, edit, and run commands themselves, under their own sandbox and |
| 111 | approval rules. They start in their own default mode, and their mode picker in the model options |
| 112 | menu switches it. How far an agent is confined depends on that mode and the agent's own settings, |
| 113 | so use its sandbox and permission controls when that matters. |
| 114 | |
| 115 | When a registry agent asks for approval, T3 Code answers by the thread's |
| 116 | [permission mode](./permission-modes.md): **Supervised** shows the request in the conversation, |
| 117 | **Auto-accept edits** approves edits and shows the rest, and **Auto** and **Full access** approve |
| 118 | automatically. File reads and searches never wait for approval. |
| 119 | |
| 120 | Devin runs its commands through T3 Code's terminals, and those commands follow the thread's |
| 121 | permission mode. For other agents, T3 Code shows the command, output, and exit status the agent |
| 122 | reports but does not run or stop their terminals. |
| 123 | |
| 124 | Registry agents can schedule work and use T3's MCP tools. Child-task presentation depends on what |
| 125 | the agent exposes: ACP has no portable native subagent-lineage contract, so richer delegation views |
| 126 | remain agent-specific. |
| 127 | |
| 128 | ## Checkpoints |
| 129 | |
| 130 | Checkpoint rollback restores your files as usual. ACP agents cannot rewind their own conversation, |
| 131 | so the next turn after a rollback starts a fresh agent session and T3 Code sends it the kept |
| 132 | conversation, without the turns you rolled back. |
| 133 | |
| 134 | Registry instances are not used for T3's app-owned text generation, such as thread titles, commit |
| 135 | messages, branch names, or pull request descriptions. Configure a text-generation-capable provider |
| 136 | for those actions. |
| 137 | |
| 138 | ## Advanced configuration |
| 139 | |
| 140 | - **Executable override** runs an existing local executable instead of the managed distribution, |
| 141 | keeping the registry-declared arguments and environment. |
| 142 | - **Authentication method** picks a specific method when the agent advertises more than one. |
| 143 | - **Custom models** adds model IDs the agent does not report. |