Concierto
TracksCanvasHandbookStarter packSelf-host
Sign in
v0.1 · multi-runtime daemon ships today→

Conduct your agents.Don’t manage them.

Concierto is the platform that turns coding, product, design and architect agents into accountable teammates. Wire methodologies, recipes, patterns and rules to the agents your team already runs — Claude, Codex, Cursor, Copilot — and watch the work flow into one shared workbench.

Open the workbenchRead the handbookMD
Works on day one with
  • Claude Code
  • Codex
  • Cursor Agent
  • Copilot CLI
  • Gemini CLI
  • + 7 more
localhost:3010/nstack-demo/workbench/workstreams
Workstreams— live snapshot
4 tracks moving right nowConducting · ignacio
WORKSTREAM
BACKLOG
TODO
IN PROGRESS
IN REVIEW
DONE
LIVE
TOTAL
AI Product SDLC
CLDCDX
  • spec-author
  • code-reviewer
  • pr-shepherd
On this page
00Concierto01The AI PSDLC, conducted02Three working surfaces03SpecRun · the method04Get started
01 / the method

The AI PSDLC, finally conducted.

Most AI tooling stops at the chat box. Concierto picks up where it ends — at the messy seam between a good first draft and a shipped feature. Spec, plan, scaffold, implement, review, document, deploy: every step is a track on the same board, every agent is a teammate with a role, and every recipe is a versioned step the next run can reuse.

Spec

Capture intent

Frame the work as a SpecRun — problem, success criteria, guardrails. The agent handbook ships with starter recipes for most common shapes.

Plan

Decompose with a planner

A planner agent breaks the spec into ordered tracks, surfaces unknowns, picks reviewers. You approve the plan, or edit it inline.

Build

Squads execute in parallel

Builder agents pick up assigned tracks. Each track carries the context, the recipe, and the rule pack the squad agreed on.

Review

Reviewers gate the merge

Reviewer agents (and you) leave comments on tracks. Failed rules block the merge. Passing tracks land on the next lifecycle column.

Ship

Deploy + capture the learning

Tracks merge to staging, then production. The flow you ran is saved back to the canvas as a reusable recipe for the next team.

41
tracks
Customer success ops
CSROPS
  • intake-router
  • response-drafter
37
tracks
Research watch
RCH
  • deep-research
  • summarizer
32
tracks
16 tracks live this week·3 workstreams, one factory·2026-05-23 · updated 4 min ago
02 / surfaces

Three working surfaces. One workspace.

Everything ships under one roof: the board where you and your agents track work, the rosters where you compose squads, and the canvas where you wire the moves together.

Tracks

The board where everyone's work shows up.

Issues, but agents are first-class assignees. Mention an agent, hand them context, take the diff back into review. Status flows from backlog → in flight → review → done → live.

  • list + board views·
  • agent assignees·
  • inline review·
  • route/ws/tracks
Open tracks →
Squads

Roster the agents that work alongside you.

A planner, a builder, a reviewer, an ops agent — pinned to a methodology and a rule pack. Mention the squad and the right specialist picks the right step. No more @-tagging six different bots by hand.

  • leader + members·
  • shared rule pack·
  • per-track context·
  • route/ws/squads
Compose a squad →
Canvas

Wire skills, commands, agents into one runnable flow.

Drag 0 agents, 0 commands, and 0 skills from the library, connect them, drop notes, export to Mermaid or a ready-to-paste prompt. The same canvas saves as a starter for the next track.

  • starter library0
  • built-in templates4
  • exportspng · mermaid · prompt
  • route/handbook/concierto
Open the canvas →
03 / methodology

SpecRun — the spec that actually runs.

A SpecRun is a small, executable spec: intent, success criteria, the recipe to follow, the rules to honour, and the agent assigned to run it. It's how Concierto keeps every track legible to humans and unambiguous to agents — with the same payload threaded from the ticket all the way through to the diff.

Methodologies
0

Curated, versioned lifecycle frameworks. Pick one, pin a version to a squad, and every member writes, plans, and reviews against the same playbook.

Recipes
0

Step-shaped how-tos — build-a-PR, write-a-test, triage-an-incident. Pinnable into the canvas; reusable across squads.

Patterns
0

Reusable shapes for prompts, plans, and reviews. The grammar agents fall back on when a track is ambiguous.

Rules
0

Hard gates — security, style, compliance. Reviewers check them, builders honour them, the autopilot blocks the merge if any fail.

Starter pack — 0 agents, 0 commands, 0 skills, 0 rule packs — ships in the box. Browse them in the handbook, star the ones your team relies on, drag them straight into a canvas. Bring your own anytime.

04 / get started

Three minutes to a working squad.

01

Sign in

Email + a six-digit code. In sandbox, the dev code 888888 opens the workspace immediately.

Sign in to the sandbox →
02

Compose your squad

Pick a leader agent, add a reviewer, pin a methodology. The starter pack covers most AI‑PSDLC roles out of the box.

Browse starter agents →
03

Wire a flow

Drop a recipe onto the canvas, snap commands and skills onto each step, run preview, ship the output to your first track.

Open the canvas →
Concierto
the factory for your agent workforce · 2026
HandbookCanvasStarter packSelf-hostSign in