Which approach fits you?
Three comparison pages on this site place driftless next to three different takes on spec-driven development: Kiro, AWS's agentic IDE; drft, a governance layer for external coding agents; and Spec Kit, GitHub's open source toolkit. This entry collects the differentiators those pages establish and restates their fit guidance in one place. Every competitor fact below stays exactly as the comparison pages state it.
The differences at a glance
| Differentiator | driftless | Kiro | drft | Spec Kit |
|---|---|---|---|---|
| What it is | Managed delivery platform hosted at app.godriftless.ai | AWS's agentic IDE built around spec-driven development | Governance layer around external coding agents, by its own description | Open source toolkit (MIT) installed into your repository |
| Spec gates | Platform-enforced on every task; a failed gate stops the work in place | Automated reasoning checks requirements for contradictions and gaps, as analysis inside the IDE | Human approval gates at every stage | Opt-in additions you invoke when you need extra quality gates |
| Requirements shaping | BA Agent converts conviction into paired PRDs and tech specs | Prompts become requirements, designs, and sequenced task lists | Refinement agents structure rough PM specs into versioned, machine-readable requirements | Skills run in your agent's chat: constitution, specify, plan, tasks, implement, converge |
| Agents | Persistent BA and coding agents included and orchestrated by the platform | Parallel agents in IDE, CLI, web, and cloud sessions | No coding agents of its own; Claude Code, Cursor, Codex, or Gemini codes through MCP | Your AI coding agent runs the process; setup needs Python 3.11+, uv, and a supported agent |
| Spec-to-task traceability | Bidirectional: requirement to task, task to requirement | Not documented on kiro.dev as of September 24, 2026 | Versioned specs and plans tied to the approved spec | Not documented in the Spec Kit README as of September 24, 2026 |
| Pricing | Per seat, flat monthly price with no credit metering; free tier: 1 user, 2 agents (BA + Coding), BYOK inference; Pro is $100/seat/mo for the first 100 seats and $70/seat after; Enterprise is custom | Per user, credit-metered; Pro is $20/user/mo with 1,000 credits, higher tiers run to Power at $200 with 10,000; add-on credits cost $0.04 each and unused monthly credits do not roll over | Free tier today: 1 workspace, 1 project, 3 team members, unlimited specs; a Pro tier is listed as coming soon | Free and MIT licensed; you pay for your own coding agent and model usage; self-managed |
Who each approach fits
- Kiro: choose driftless when you want the delivery cycle itself managed and enforced; choose Kiro when the work you need is inside an editor.
- drft: with driftless you get shaping, execution, tracking, and verification in one place; with drft you add governance on top of agents you already run.
- Spec Kit: Spec Kit gives you the process artifacts; driftless gives you the process plus the platform that runs it. Spec Kit costs nothing and runs entirely on your own agent and models, while driftless is fully managed with hosting, agents, and analytics included.
Do Kiro, drft, and Spec Kit enforce spec gates?
Kiro checks requirements for contradictions and gaps with automated reasoning, and those checks run as analysis inside the IDE. Spec Kit's quality gates are opt-in additions you invoke when you need extra quality gates, per the project README. drft adds human approval gates at every stage of its draft, refine, approve, plan, implement, verify cycle. driftless runs spec gates as part of the platform workflow on every task, so a failed gate stops the work in place.
Does driftless run the agents itself?
Yes. driftless includes a BA agent for shaping specs and coding agents that the orchestrator directs against task directives, with delivery metrics and per-model cost tracking. drft connects to coding agents you already run, such as Claude Code, Cursor, Codex, or Gemini, through MCP.