BoltOps is an AI-first sprint management platform. You describe what to build, AI agents generate and verify the code, humans review and approve, and the whole team ships faster. Here's how every piece fits together.
Define what to build (Intent), review AI output (Review), approve changes (Iterate), and deploy (Ship). You're always in the driver's seat.
Claude Code generates code. QA Agent verifies it. Docs Agent writes release notes. All managed by the Paperclip orchestration layer with budget controls.
Clients, managers, or teammates who review deliverables via Magic Links. No account needed. They give severity-rated feedback that feeds the iterate cycle.
Every Bolt flows through the same 6-stage pipeline below. Stages colored blue are AI-powered.
Every Bolt (unit of work) moves through these six stages. Three are AI-powered, two are human-driven, and one is collaborative.
Performed by: You (Product / Engineering)
Describe what you want to build in a structured Intent Document — title, priority, project, and acceptance criteria.
Hit "Generate with Claude" and AI auto-fills user stories, edge cases, and acceptance criteria from your title alone.
Tip: Think of an Intent Document as a mini-PRD. The better your title, the better Claude’s output.
Performed by: AI Agents (Claude Code)
When a Bolt enters this stage, a Claude Code session auto-dispatches. The agent reads your Intent Document + CLAUDE.md guidelines and generates code.
Fully automated. Claude Code writes implementation code, scaffolds tests, and captures every prompt, output, and token cost.
Tip: You stay in control — set budget limits and approval gates so no agent runs without your say-so.
Performed by: AI Agents (QA Agent)
Automated verification: the QA agent runs tests, linting, type checks, and security scans against the generated code.
The QA agent validates outputs against your acceptance criteria from the Intent Document. Issues are flagged with severity ratings.
Tip: Verification results appear on the Bolt detail page — green means ready for human review.
Performed by: You (Human Review Gate)
A human reviews the AI-generated work. You can approve, request changes, or share a Magic Link with external stakeholders for feedback.
No AI acts here — this is your control point. But you can generate a Magic Link so clients or teammates review without logging in.
Tip: Magic Links support severity-rated feedback: Must Fix, Should Fix, Nice to Have, and Discussion.
Performed by: AI Agents + You
Review feedback flows back to the Code Agent. Claude reads the feedback, makes targeted changes, and re-submits. You review again.
The Code Agent automatically reads "Must Fix" and "Should Fix" items and generates patches. You approve each change.
Tip: Most Bolts go through 1-2 iteration cycles. The feedback loop is intentionally tight.
Performed by: You + Docs Agent
You handle the final deploy. The Docs Agent auto-generates release notes and changelog entries from the Bolt’s history.
Release notes, API docs, and changelog entries are generated automatically from the intent doc and code diffs.
Tip: Group multiple Bolts into a Release Train with dependency tracking and scope gates.
These are the core concepts you'll encounter. Understanding them makes everything else click.
A unit of work — like a ticket in Jira, but designed for AI. Every Bolt flows through the 6-stage pipeline from Intent to Ship.
A structured spec attached to every Bolt. Contains the title, priority, user stories, acceptance criteria, and AI-generated details.
The multi-agent orchestration layer. Dispatches specialized AI agents (Code, QA, Docs, Review) with budget controls and approval gates.
Per-project AI coding guidelines. Version-controlled, with section ownership and PR-style reviews. AI agents follow these rules when generating code.
Shareable URLs that let anyone review a Bolt’s deliverables without an account. Configurable expiration, view limits, and severity-rated feedback.
Analytics dashboard showing cycle time, throughput, token usage, and per-agent cost breakdowns with executive-ready reports.
The key principle: AI does the heavy lifting. You make the decisions. Every AI action has a human approval gate. Nothing ships without your sign-off.
Follow these six steps to go from zero to your first shipped feature.
Sign up free — no credit card required. You get 3 seats and 25 Bolts/month on the Starter plan.
Sign Up FreeA Project is a workspace for your team. Connect it to a GitHub repo and BoltOps will sync your CLAUDE.md automatically.
Click "New Bolt", give it a title like "Add user authentication with SSO", set the priority, and hit "Generate with Claude" to auto-fill the details.
Drag the Bolt to Generate. Claude Code auto-dispatches a coding session. Watch real-time progress, token usage, and cost on the Bolt detail page.
Review the generated code. Share a Magic Link with stakeholders for feedback. The iterate cycle turns feedback into code changes automatically.
Drag to Ship. The Docs Agent generates release notes. Group Bolts into Release Trains for coordinated deployments.
No. BoltOps is designed for product managers, engineering leads, and developers alike. You describe what you want in plain language. AI handles the implementation. Developers can dive into the generated code if they want to.
Nothing ships without your approval. The Review stage is a mandatory human gate. You (or your stakeholders via Magic Links) review everything before it moves forward. If something is wrong, the Iterate stage sends it back to AI for fixes.
Paperclip budget controls let you set per-agent spending limits, approval gates for expensive operations, and org-level caps. You can see the cost of every session in real-time on the Bolt detail page.
Yes. The Starter plan gives you the full 6-stage pipeline and Kanban board without AI agent integrations. You can use it as a lightweight project management tool and add AI later.
Jira is a general-purpose project management tool. BoltOps is purpose-built for AI-first development — it has built-in AI agent orchestration, per-agent cost tracking, CLAUDE.md governance, and Magic Links. The 6-stage pipeline is opinionated by design.
It's a version-controlled file in your repo that defines how AI agents should write code for that project — coding standards, architecture patterns, forbidden patterns, and more. AI agents read it before generating code.
Start free. No credit card. 3 seats. 25 Bolts/month. Your first feature ships today.