AI Productivity

Agent-to-Agent Reads

Q: How does Agent-to-Agent Reads work in 1DevTool?

Step 1: Link two terminals — Reads travel over an existing link, so the agents must already be connected. Sending and asking work with no read permission at all. Step 2: Grant what it may read — Approve read-transcript, read-transcript-full, read-screen or read-artifact separately. Full transcript adds to the bounded grant, it never replaces it. Step 3: Read the disclosure scope — The dialog names every agent whose content the grant could reach, warns that screen reads return keys and .env output verbatim, and says revoking cannot recall what was already copied. Step 4: The agent looks instead of asking — It pulls the peer’s transcript, screen or published file and uses the answer in the same turn. The peer keeps working — nothing queues behind its current turn.

A linked agent can pull its peer's work directly — a bounded slice of its transcript, the full transcript, the terminal screen as rendered, or a file the peer published — and use the answer in the same turn. The peer is never interrupted, and nothing is readable until you grant it per link.

Download 1DevTool All Features

Why You Need This

How this feature boosts your daily productivity

Linked agents could already send to each other and ask each other, but neither one could look. So when the agent about to write a fix needed to know what the agent that traced the bug had just learned, you became the wire — scrolling one terminal, copying, switching panes, pasting — or you sent a message and waited for the other agent to finish its turn before it answered. Reads remove both. The requesting agent gets what it needs inside its own turn, and the peer keeps working.

How It Works

Get started in 4 simple steps

Link two terminals

Reads travel over an existing link, so the agents must already be connected. Sending and asking work with no read permission at all.

Grant what it may read

Approve read-transcript, read-transcript-full, read-screen or read-artifact separately. Full transcript adds to the bounded grant, it never replaces it.

Read the disclosure scope

The dialog names every agent whose content the grant could reach, warns that screen reads return keys and .env output verbatim, and says revoking cannot recall what was already copied.

The agent looks instead of asking

It pulls the peer’s transcript, screen or published file and uses the answer in the same turn. The peer keeps working — nothing queues behind its current turn.

Without 1DevTool

The manual way vs. the 1DevTool way

The Hard Way

# Agent A traced the bug

# Agent B is about to write the fix

# Scroll A’s terminal, find the part that matters

# Copy, switch panes, paste

# Or message A and wait for its turn to end

# You are the wire between two agents

The 1DevTool Way

B reads A’s transcript in its own turn.

A is never interrupted.

Four grants, each off until you give it.

On macOS, every read proves who is asking.

Key Capabilities

Everything included with Agent-to-Agent Reads

Read a peer’s recent transcript without interrupting its current turn

Request the full thread when recent context is not enough

Read the peer’s terminal screen exactly as it is rendered

Pick up a file or result a peer published for one specific link

Ask whether anything has changed since the last read instead of re-reading everything

Grant read-transcript, read-transcript-full, read-screen and read-artifact independently

See every agent a grant would expose before approving it

Revoke a grant at any time — a read already in flight is cut off before it returns

Have each read verified against macOS before it is answered

Why It Matters

The productivity impact of having ai productivity built into your workspace

Searchable Prompt Library

Every prompt you've ever sent is captured and searchable. Find and reuse your best AI interactions.

Visual AI Communication

Annotate screenshots and embed them in prompts. Show your AI agent exactly what you see.

Session Knowledge Base

AI conversations become searchable history. Find how you solved a similar problem last week.

Power Navigation

Command palette gives you instant access to any feature. Navigate 1DevTool at the speed of thought.

More in AI Productivity

Other features in this category you might find useful

Prompt History

Search prompts and sticky notes from Remote Control while away from the desktop. Switch between This project and All projects, filter prompts by agent or notes by source, copy entries, delete with Operator access, and sync local prompt history.

Resume UI

Read any past AI session before you resume it. Resume AI Sessions is a split view with a searchable list of every Claude Code, Codex, Gemini CLI and other agent session on the left and a live preview of the project, branch, tool calls and conversation on the right, with replies rendered as Markdown, with Resume, bypass permissions and an editable command one click away.

Combine Sessions

Merge several past AI sessions into one opening prompt, or hand a single session to a different agent. Tick sessions in Resume AI Sessions, see a token estimate of the context they add, and send the result to a new chat with any agent or preset or to a terminal that is already open.

AI Diff Review Panel

Track every file modification any AI agent makes — Claude Code, Codex, Gemini, Amp, custom — and review, accept, or revert their work per file from a dedicated panel with inline diff overlays in the editor.

The RouterPrevious All Features Agents Home on Your PhoneNext

Ready to boost your workflow?

Download 1DevTool for free and experience Agent-to-Agent Reads along with 335 other features — all in one app.

Download for Free View Changelog