Teams of agents, not one chat
Use powerful models for hard work and lower-cost models where they fit. Assign the right agent to the task, then keep the work connected.
AuraPunk ADE · Agentic Development Environment
Plan from Kanban cards, run multiple agents in parallel, and retain the context that matters. AuraPunk ADE brings the workflow, memory, and integrations into one operating surface.
Who AuraPunk ADE is for
A ready environment for using the right model, agent, and level of oversight for every task.
Use powerful models for hard work and lower-cost models where they fit. Assign the right agent to the task, then keep the work connected.
Run planning, research, implementation, and review at the same time. Cards, chats, decisions, and audit logs remain attached to the project.
Reply when an agent needs an answer, follow progress, or review a result from mobile—without needing to remain at the terminal.
Graph, semantic, and vector memory; SPECs; agent orchestration; Jev; and Laya are composed in one environment, ready for the next way of building.
Installation & downloads
A native development environment for parallel coding agents, persistent project memory, and optional human review when it matters.
AuraPunk Mobile lets you manage models and providers, open workspaces, follow cards and chats, and send work to your connected Desktop or Cloud instance from the same account.
Standalone disk image (.dmg). Drag the app to Applications and start building your projects.
Because AuraPunk ADE uses an independent signature, macOS Gatekeeper may ask you to confirm the first launch after a browser download:
Open the downloaded .dmg file and drag AuraPunk ADE to Applications.
If macOS shows “Apple cannot verify the developer”, open System Settings → Privacy & Security → Open Anyway.
Or allow it instantly from Terminal:
xattr -cr "/Applications/Aurapunk IDE.app"The development workflow
Turn intent into an executable card. Scope, acceptance criteria, code, logs, decisions, and review stay attached to the work—not scattered across chats.
↗Dispatch planning, coding, research, and review agents in parallel. Each agent has a defined role while AuraPunk preserves the shared project thread.
↗Stable context prefixes, prompt reuse, and automatic compaction help more requests land on cache—so your token budget goes into new work.
↗Persistent workspace history
Every chat, decision, agent run, code change, and review stays connected to the project. Return tomorrow, switch devices, or reopen a card and continue from exactly where you stopped—with a complete, auditable record of what happened.
Explore project memory ↓↺ Resume from the last verified state
Memory, by approach
AuraPunk ADE combines SPECs, semantic memory, and vector memory. Agents retrieve the right kind of context for the task, while scopes and provenance keep it useful.
Explore the memory approaches ↓Three memory approaches
Structured project context keeps active plans, decisions, and execution state available to the agents that need to move the work forward.
Meaning-aware retrieval reconnects an agent to relevant decisions, patterns, and prior work even when the wording changes.
Fast similarity retrieval gives agents a precise path back to code, documents, and knowledge held across the project.
Your temple for agentic technology
The newest agentic work is already here, without the setup. Laya and Jev beside Obsidian, semantic, and graph memory — AuraPunk is the easiest way to use all of it, in one surface, the same way every day.
System-1 inference for the high-frequency decisions inside a workflow. Fast, deliberate, and already wired into the cards that need it.
Typed judgment where the workflow needs one: natural-language signals become decisions your system can act on, without hand-rolling the plumbing.
Your vault becomes durable agent context. Notes you already keep are readable by the agents working the project, so knowledge stops living in a second place.
Meaning-aware retrieval that reconnects an agent to the right decision, pattern, and prior work even when the wording changes.
Every card, chat, decision, and edit kept as a relation — so context compounds over the project instead of being re-explained in every new chat.

High-end integrations
Use Jev where a fast, typed judgment belongs in the workflow—turning natural-language signals into decisions your system can act on.
+Laya adds System-1 inference for lightweight decision tasks, so high-frequency workflow decisions remain responsive and deliberate.
+The ADE connects cards, agents, memory systems, and model integrations without turning your development process into a collection of disconnected tools.
+Free Multi-Agent Development Environment
AuraPunk ADE's full source is open on GitHub, free to use with every feature enabled—including its multi-agent workflow and self-hosted memory stack. The Personal plan swaps self-hosting for a managed cloud instance and cloud-hosted memory, fully configured and ready to go.
View source on GitHub ↗The first workspace is forming
AuraPunk is for people who want more from their agents—and more agency over the result.
Start a conversation ↗Follow us on X@aurapunkdevAURA / 001