AI coding agents have changed how software gets built. Claude Code, Codex, Gemini CLI — each one can reason about code, make changes, and iterate on problems that would take a human developer significantly longer. But as soon as you start using more than one of them, a new problem emerges.
The Problem
If you work with multiple AI coding agents, your workflow probably looks something like this: one terminal tab running Claude Code on a backend refactor, another tab with Codex generating tests, and a third with Gemini exploring an integration approach. You alt-tab between them, trying to keep track of which agent changed which files, whether their modifications conflict, and what state each session is actually in.
There is no unified view. No way to see at a glance what every agent is doing, which files they have touched, or how their work relates to each other. You are the orchestration layer, and your tools for the job are terminal tabs and your own short-term memory.
This does not scale. As AI agents become more capable and developers run more of them in parallel, the coordination overhead grows faster than the productivity gains. The missing piece is not a better agent — it is a better way to manage all of them.
What Is Glueprint
Glueprint is an Agentic Development Environment (ADE) — the IDE reimagined for agents. Where the IDE was built for you to write code on one machine, the ADE is built for you to direct the agents that write it — from anywhere. It is not another AI assistant. It is the layer that sits above your agents and below your workflow, bringing every agent, every host, and every development environment you work in into one place.
Built with Tauri 2.0 and Rust, Glueprint connects to your running agent processes, parses their output in real time, and presents everything in a structured interface. You see every session, every file change, every git diff, and every task dependency in one place — on your desktop, in any browser, or from your phone. When agents finish work, Glueprint can chain the next task automatically.
Think of it as the IDE for the agentic era: a single environment that runs the agents writing your code and surfaces what they are doing wherever you are.
Key Capabilities
Real-Time Session Streaming
Glueprint attaches to agent processes and parses stdout in real time. Raw terminal output is transformed into rich, structured message types — user prompts, assistant responses, thinking blocks, tool invocations, and tool results — using discriminated unions with type guards for reliable narrowing. You see exactly what an agent is doing at every moment, rendered with the clarity of a purpose-built interface rather than a wall of terminal text.
File Browser
A built-in file browser lets you navigate your project directories without leaving the app. As agents create, modify, or delete files, the changes appear immediately. You always know what the current state of your working tree looks like, even when multiple agents are making changes concurrently.
Git Integration
Glueprint tracks staged and unstaged changes across your repository. View diffs inline, see which agent introduced which change, and commit directly from the app. When agents are working on related areas of the codebase, you catch conflicts before they compound.
Terminals
Glueprint ships a full xterm.js terminal on every surface — desktop, portal, and mobile — with ANSI colors, scrollback, copy-paste, and splittable tab grids. Sessions live on the agent host, so closing the desktop app doesn’t kill your shell; reopen from your phone and pick up mid-stream. Output rides the same encrypted relay as your sessions, so the server sees envelopes, not your commands.
Tasks and Boards
Glueprint organizes work on boards — one per project by default, one per assistant team, and personal boards for individual focus. Tasks carry assignment, due dates, and statuses governed by a Task Flow Definition: a state machine that defines which transitions are allowed and which roles can make them. The same board renders as a list, kanban, swimlane, or queue, with full-text search, overdue surfacing, and bulk operations. This is where Glueprint moves from monitoring tool to orchestration platform — not by chaining tasks blindly, but by codifying who can move what and when. Learn more in our guide to multi-agent workflows.
AI Assistants
Spin up long-running AI assistants — each with its own persona, constitution, memory, and task queue. Assistants run on the harness of your choice (Claude Code, Codex, Gemini, or any of the 33 ACP-compatible harnesses in the registry), wake on cron-driven schedules, work through their queue, journal their progress, and report back through a dedicated chat channel. Group assistants into teams with role-gated workflows to coordinate multi-agent work across hosts, and govern each one with permission tiers, tool filters, budgets, and quiet hours.
Unlike standalone AI assistants like OpenClaw or Hermes, Glueprint’s assistants are harness-agnostic and run alongside your sessions, files, and git — inside the same encrypted environment. Pick the model. Pick the agent runtime. Govern them all from one place.
Team Chat
Talk to your team, your assistants, and the agent in any live session — all inside Glueprint, all on the same end-to-end encrypted relay. The mobile app brings those same conversations to your phone, so you can approve a tool call, redirect an agent, or hand off work without ever leaving the trust boundary of your org. There is no need to pipe prompts, source code, or agent output through Slack, Discord, or Teams; the server sees ciphertext, not your messages.
Why Tauri
Glueprint is built on Tauri 2.0, and the choice is deliberate. Tauri apps ship as native binaries with a fraction of the memory footprint of Electron-based alternatives. On a typical machine, Glueprint uses roughly 50-80 MB of RAM — a meaningful difference when you are already running multiple agent processes that each consume significant resources.
The architecture splits cleanly: the Rust backend handles everything performance-sensitive — file I/O, process management, encryption, WebSocket connections, and atomic writes. The React frontend handles rendering and interaction. Communication flows through Tauri’s IPC layer with typed commands, so the boundary between frontend and backend is well-defined and type-safe.
Data is stored locally in ~/.glueprint/ using flat files. Configuration and work items use YAML. Sessions and audit logs use JSONL with atomic writes for crash safety. There is no local database to configure, no migrations to run, and no server to start. Your data lives on your machine in formats you can inspect with any text editor.
Optional Cloud Relay
For teams that need remote visibility — reviewing an agent’s progress from a phone, sharing a session with a teammate, or monitoring a long-running pipeline from another machine — Glueprint offers an optional cloud relay.
The relay is built on a zero-knowledge encryption model. Every event is encrypted client-side with AES-256-GCM before it leaves your machine. The relay server routes encrypted envelopes between clients but never sees plaintext content. Decryption keys are derived client-side from a device-held secret using PBKDF2-SHA256 with 600,000 iterations, so only your clients can read the data.
If a client disconnects and reconnects, the protocol recovers any missed events automatically — no full resyncs, no data loss.
The cloud relay is entirely optional. Glueprint works fully offline, and no data leaves your machine unless you explicitly configure a relay connection.
Get Started
Glueprint is available today. Download the desktop app, install the CLI daemon on a server, open the web portal, or grab the mobile app — the four surfaces share one encrypted relay. The desktop app runs on macOS, Windows, and Linux. Check out the getting started guide for a step-by-step walkthrough.
Getting started takes less than a minute:
- Download and install the app for your platform.
- Enter your activation key on the registration screen.
- Start your AI coding agents as you normally would — Claude Code, Codex, Gemini, or any ACP-compatible harness.
- Connect and orchestrate from desktop, browser, or phone.
Claude Code, Codex, and Gemini ship as first-class adapters; any ACP-compatible (Agent Client Protocol) harness plugs in alongside them — 33 in the registry today. There is no vendor lock-in and no proprietary protocol your agents need to support.
We built Glueprint because we ran into the tab-switching problem ourselves, and we suspect most developers using AI agents have hit the same wall. The agents are powerful. The missing piece was an environment that runs all of them together. That is what the ADE provides.
Visit glueprint.ai to download the app and start orchestrating.