Interactive control

This tutorial walks you through Untether’s interactive permission system — approving, denying, and shaping agent actions from Telegram on whatever device is in your hand. Stay in control while you’re away from the terminal.

What you’ll learn: How to control Claude Code’s actions in real time with Telegram buttons, how to request and review a plan before execution, and how to answer agent questions from anywhere.

Claude Code only

Interactive approval buttons (Approve / Deny / Pause & Outline Plan) are a Claude Code feature. Other engines run non-interactively. Codex CLI has a pre-run approval policy in /config (full auto vs safe = read-only sandbox) and the deprecated Gemini CLI has an approval mode, but neither has per-tool interactive buttons.

1. Understand permission modes#

Untether offers four permission modes that control how much oversight you have:

For this tutorial, use Plan mode so you can see the plan approval flow.

2. Enable plan mode#

Open your Telegram chat with the bot and send:

/planmode on

Untether

permission mode on for this chat: plan mode: Claude plans without editing files, and you approve the plan before changes start.
applies from your next message (--permission-mode plan). Until then, the current run and any background wake-ups keep the old mode.

The bot confirms that plan mode is now active. This setting is stored per chat and persists across sessions; it takes effect from your next message.

3. Send a task#

Send Claude Code a task that will require file changes:

add a comment to the top of README.md explaining what this project does

Claude Code starts working and you’ll see a progress message stream in.

4. See approval buttons#

When Claude Code needs your go-ahead, Untether intercepts the request and shows you what’s about to happen. In plan mode the main checkpoint is the plan approval (step 7). Individual tool calls come to you too when Diff preview is on in /config, and with /planmode off for anything beyond file edits (a shell command shows as $ <command>). A file edit looks like this:

The message includes:

• Request type and tool name with key parameters (e.g. file path, command)
• Diff preview — 📝 file path, removed lines (- old) and added lines (+ new) in a monospace diff block
• Buttons: ✅ Approve and ❌ Deny. A plan approval (Claude Code asking to leave plan mode) reads ✅ Approve Plan instead and adds a third button, 📋 Pause & Outline Plan (step 7)

Your phone will also buzz with a push notification so you don’t miss it.

5. Approve a tool call#

Tap Approve to let Claude Code proceed with the action. The button clears instantly — no spinner, no waiting. Claude Code continues with its work.

Untether

working · claude · 8s · step 2

✓ Read README.md
▸ Edit README.md

You may see several approval requests in a row as Claude Code works through multiple steps.

6. Deny a tool call#

If something doesn’t look right, tap Deny instead. Claude Code receives a denial message explaining that you’ve blocked the action and asking it to communicate via visible text instead.

Untether

I was going to add a comment to the top of README.md with a project description. Would you prefer I explain what I had in mind first, or should I take a different approach?

This is useful when you want Claude Code to explain its reasoning before making changes. After denying, Claude Code will typically describe what it was trying to do and ask for guidance.

7. Use “Pause & Outline Plan”#

In plan mode, when Claude Code has a plan and tries to exit plan mode (the move from planning to execution), you get ✅ Approve Plan, ❌ Deny and a third button — 📋 Pause & Outline Plan — which is the most powerful. The message also says what approving does: Claude carries out the plan without further prompts, and plan mode resumes when that reply ends (or once any background agents it started have finished).

Tap it to require Claude Code to write a comprehensive plan as a visible message before doing anything. The plan must include:

1. Every file to be created or modified (full paths)
2. What specific changes will be made in each file
3. The execution order and any key decisions or risks

and run to at least 15 lines of visible text. An outline that is too short (under about 200 characters) is rejected and Claude Code is asked again.

The outline renders as formatted Telegram text — headings, bold, code blocks, and lists display properly instead of raw markdown:

Untether

Here’s my plan:

1. Read README.md to understand current content
2. Edit README.md to add a project description comment at line 1
3. Verify the comment is correctly formatted

Files to modify: README.md

After Claude Code writes the outline, Approve Plan, Deny, and Let’s discuss buttons appear automatically on the last message of the outline — no need to scroll back up or type “approved”:

• Tap Approve Plan to let Claude Code proceed with implementation. The approval covers that reply only: your next message starts in plan mode again
• Tap Deny to stop Claude Code and provide different direction
• Tap Let’s discuss to talk about the plan before deciding — Claude Code will ask what you’d like to change and wait for your reply

Outline gate

After tapping “Pause & Outline Plan”, Untether holds ExitPlanMode open until Claude Code provides a readable outline — the plan is posted to the chat and the agent stays alive while you read it. If Claude Code retries without an outline, that attempt is auto-denied with an instruction to write the outline first. (Earlier versions also enforced a 30–120s escalating cooldown here — a workaround for a Claude Code v2.1.72–2.1.74 retry loop that was fixed upstream and retired in v0.35.4.)

8. Answer a question#

Sometimes Claude Code needs to ask you something — like which approach to take or what naming convention to use. When Claude Code calls AskUserQuestion, you’ll see the question in the chat with a ❓ prefix and option buttons for each choice:

Tap an option button to select your answer. Claude Code receives your choice and continues immediately.

For multi-question flows (1 of N, 2 of N), each question appears in sequence after you answer the previous one.

If none of the options fit, tap Other (type reply) and type a custom answer as text. Untether routes your reply back to Claude Code, which reads it and continues.

You

Use snake_case for all variable names

Untether routes your reply back to Claude Code, which reads it and continues.

There’s no Deny button on a question with options: if none of them fit, pick Other (type reply) and say so.

Ask mode toggle

Control whether Claude Code asks interactive questions via /config → Ask mode. When off, Claude Code proceeds with reasonable defaults instead of asking.

9. Switch to plan-auto mode#

Once you’re comfortable with how Claude Code works, you might want less interruption. Switch to plan-auto mode:

/planmode plan-auto

Untether

permission mode plan-auto for this chat: plan mode, but the plan is approved for you: no plan buttons.
applies from your next message (--permission-mode plan). Until then, the current run and any background wake-ups keep the old mode.

In plan-auto mode, tool calls (Edit, Write, Bash) are auto-approved — Claude Code works without interruption. Plan transitions are also auto-approved, so you won’t see ExitPlanMode buttons. The agent preamble still requests summaries and structured output.

/planmode auto is different: it hands approval to Claude Code’s own classifier, which runs routine work and blocks risky actions. See Plan mode.

10. Skip the plan phase#

To drop the plan phase (Claude Code’s acceptEdits mode):

/planmode off

This sets Claude Code to acceptEdits mode: no plan phase, and file edits run without buttons. Since v0.36.0, anything acceptEdits doesn’t cover (most shell commands, web fetches, MCP tools) shows Approve / Deny buttons, unless your Claude Code permissions.allow rules already allow it.

To go back to the engine’s default instead, send /planmode clear.

To check your current mode at any time:

/planmode show

Untether

permission mode: plan-auto (plan): plan mode, but the plan is approved for you: no plan buttons.

What just happened#

Key concepts:

• Permission modes control the level of oversight: on (plan mode: Claude plans without editing files, and you approve the plan before changes start), plan-auto (plan mode, but the plan is approved for you: no plan buttons), auto (Claude Code’s auto mode: a classifier approves routine actions and blocks risky ones; no plan phase) and off (acceptEdits: no plan phase; file edits and common filesystem commands run, and other tools ask you first)
• Approval buttons appear inline in Telegram when Claude Code needs permission — Approve, Deny, or Pause & Outline Plan; after an outline is written, you also get Let’s discuss to talk about the plan
• Diff previews show you exactly what will change before you approve
• “Pause & Outline Plan” forces Claude Code to write a visible plan before executing
• Plan approvals are per reply — approving a plan lets Claude Code carry it out without further prompts, then plan mode resumes for your next message
• Outline formatting — plans render as proper Telegram text with headings, bold, and lists; buttons appear on the last message; outline messages are cleaned up after you act on them
• AskUserQuestion lets you answer Claude Code’s questions with option buttons or a text reply
• Push notifications ensure you don’t miss approval requests, even from another app
• Ephemeral cleanup automatically removes button messages when the run finishes

Troubleshooting#

Approval buttons don’t appear

Check that you’re using Claude Code (/claude prefix or /agent set claude) and which permission mode is active (/planmode show). engine default with no [engines.claude] permission_mode in untether.toml (a fresh install) means Claude Code runs under its own settings with no buttons — set a mode with /planmode. In plan-auto and auto no plan buttons appear by design; tool buttons appear in Accept edits / default, or with Diff preview on. Other engines don’t support interactive approval.

Buttons appear but nothing happens when I tap them

Check your internet connection. If the tap doesn’t register, try again — Untether answers callbacks immediately so there should be no delay. A toast saying Already answered, No longer needed or This request has expired means the request was already settled (or Claude Code withdrew it), so the tap was ignored. Buttons also only work in the chat they were posted in, and in a group with approval_originator_only = true only the person who started the run can answer them.

Claude Code keeps retrying after I tap “Pause & Outline Plan”

Until Claude Code writes an outline, each ExitPlanMode retry is auto-denied with an instruction to provide the plan first — this is the outline gate. Wait for Claude Code to write the outline, then use the Approve Plan / Let’s discuss / Deny buttons that appear. (The 30–120s escalating cooldown earlier versions used here was retired in v0.35.4 — the upstream retry loop it worked around is fixed.)

I don’t get push notifications for approval requests

Make sure Telegram notifications are enabled for this chat. Untether sends a separate notification message when buttons appear, but Telegram’s notification settings control whether you see it.

Next#

Now that you can control your agent interactively, learn how to target specific repos and branches.

Projects and branches →