Long-Running Project with Memory
by NeuroSquad teamVersion 1updated Oct 2, 20261 download
The app shows everything it will install and asks once before anything runs.
The canvas
About this template
For a project that lasts weeks: decisions, conventions and gotchas go into a shared memory, and come back automatically in later sessions and to the other agent.
What's inside
- Builder (Claude Code) and Reviewer (Codex CLI), both connected to the same memory.
- Project memory (Memory card, mem0 inside, stored locally in the app's data folder): agents save facts with
memory_addand search them withmemory_search. Auto-recall is on: before each prompt, the most relevant memories are added to the turn (Claude Code, Codex, Qwen, OpenCode, Kilo, pi, omp, Gemini). Scope: this workspace. Search works by keywords out of the box; turn on search by meaning in the card's settings. - Project brief note: what the project is, its constraints, who decides what.
- Milestones checklist that the Builder ticks off as it goes.
How to use
- Fill in the Project brief, adjust the Milestones.
- Press Enter on the Builder's pasted prompt.
- Come back tomorrow: a new session starts with what was learned yesterday.
To also save facts automatically at the end of each turn, pick a fact model (OpenRouter or a local provider) in the Memory card's settings and turn on auto-capture.
More templates
See all- OfficialFeaturedCoding
Bug Hunt Loop
by NeuroSquad team1 agent · 5 cards
One agent that reproduces a bug before it touches the code, finds the root cause, proves the fix with a test and writes down what happened. What's inside Fixer (Claude Code) with…
- #debugging
- #bug-fix
- #testing
- #browser
- OfficialFeaturedCoding
Worktree Crew: Lead + Parallel Workers
by NeuroSquad team3 agents · 6 cards
A lead agent splits a big change into independent tasks and runs two workers in parallel, each in its own git worktree, so they never overwrite each other's files. What's inside…
- #multi-agent
- #git-worktree
- #refactoring
- #kanban
- OfficialFeaturedCoding
Test-First Pair
by NeuroSquad team2 agents · 4 cards
Test-driven development split between two agents: one only writes tests, the other only writes the code that makes them pass. Neither can quietly weaken the other's work. What's…
- #tdd
- #testing
- #pair-programming
- #multi-agent
Put your squad to work
Free for Windows and macOS. Bring the agents you already use.
Linux is coming later.
The first-run wizard can install an agent CLI, Git and the dictation model for you — or skip each step.