Bolt New vs Kiro
A feature-by-feature comparison of Bolt New and Kiro across 7 capability categories.
Modalities — Bolt New Wins
| Feature | Bolt New | Kiro |
|---|---|---|
Text Input | Yes. Auto populated | Yes, requires free plan. Prompts become specs (requirements, design, tasks) then implementation. |
Image Upload / Vision | No. Auto populated | Partial. Homepage mentions multimodal inputs and Figma as context. A general image-upload vision API was not separately documented. |
Image Generation | No. Auto populated | No. A coding agent, not an image model. |
Voice Mode | Partial. Auto populated | No. No conversational voice loop is documented. |
Video Understanding | No. Auto populated | No. Not a video-analysis product. |
Video Generation | No. Not a video generation product. | No. Not a video generation product. |
File Upload (PDF, DOCX) | Yes. Auto populated | Yes, requires free plan. Works on local workspaces and, on the web, GitHub/GitLab repos in cloud sandboxes. |
Code Execution | Yes. Auto populated | Yes, requires free plan. Agents implement, test, and can run headless in CI. Web sessions run in isolated sandboxes. |
Audio Input | No. Auto populated | No. No audio-file input is documented. |
Canvas / Artifact Editor | Yes. Auto populated | Unknown. Overlapping columns disagreed (partial vs yes); left unverified rather than copying a no. |
Screen Sharing / Computer Use | No. Auto populated | No. No computer-use desktop takeover is documented. Agents edit the workspace. |
Memory & Context — Tie
| Feature | Bolt New | Kiro |
|---|---|---|
Persistent Memory | Yes. Auto populated | Yes, requires free plan. Steering files (.kiro/steering/, ~/.kiro/steering/, AGENTS.md) persist conventions across sessions and surfaces. |
Large Context Window | Yes. Auto populated | Unknown. Uses multiple FMs via Bedrock/Auto. No token window is published on the pricing page. |
Custom Instructions | No. Auto populated | Yes, requires free plan. Steering markdown (always / fileMatch / manual) plus AGENTS.md. |
Conversation History | Yes. Auto populated | Yes, requires free plan. IDE, CLI, web, and Crew keep sessions. Web sandboxes keep running after you close the laptop. |
Integrations — Kiro Wins
| Feature | Bolt New | Kiro |
|---|---|---|
Web Search / Browsing | Yes. Auto populated | Unknown. MCP can pull tools; a first-party live-web search toggle was not confirmed. |
Plugins / Extensions | No. Auto populated | Yes, requires free plan. MCP, Agent Client Protocol, Open VSX extensions, Skills.md, and 500+ CLI tools. |
API Access | Partial. Auto populated | Partial. Headless CLI in CI/CD and ACP into other IDEs. That is not a public REST generate API. |
Zapier Integration | No. Auto populated | Unknown. No first-party Zapier app was checked. |
Custom GPTs / Bots | No. Auto populated | Yes, requires free plan. Custom agents plus steering; ACP into other tools. Kiro Crew is the open-source coordinator. |
Real-time Collaboration | No. Auto populated | Partial. Team billing and shared steering. Live multiplayer cursors were not documented. |
Live Social Data Integration | No. Not a live social-firehose product. | |
Live Citation Mapping | No. Not a cited-search product. |
Developer — Kiro Wins
| Feature | Bolt New | Kiro |
|---|---|---|
Function Calling | No. Auto populated | Yes, requires free plan. Agents take actions (implement, test, hooks, MCP tools), not chat-only replies. |
Streaming Responses | Yes. Auto populated | Unknown. Token streaming is not documented as an API contract. |
Fine-Tuning | No. Auto populated | No. You pick models (Auto, Claude, GPT, open-weight). No custom-weight training product. |
Embeddings API | No. Auto populated | No. Not an embeddings API. |
Structured Output (JSON) | No. Auto populated | Partial. Specs and property-based tests are structured artifacts. Not a JSON-schema chat API. |
Reasoning / Thinking Mode | No. Auto populated | Unknown. Overlapping columns disagreed (yes vs partial); left unverified rather than copying a no. |
Local Execution | Yes, requires free plan. IDE and Crew run on your machine; web sandboxes are cloud. Models are hosted. | |
Open Weights | Partial. You can select open-weight models (Qwen, DeepSeek, MiniMax, GLM). The Kiro product itself is not an open-weight release except Crew. | |
Multi-Agent Orchestration | Yes, requires free plan. Parallel agents implement tasks; hooks fire on IDE events; Crew coordinates tools. |
Platforms — Kiro Wins
| Feature | Bolt New | Kiro |
|---|---|---|
Web App | Yes. Auto populated | Yes, requires plus plan. Kiro on the web (preview) is for paid Pro+ tiers; sandboxes keep running in the cloud. |
iOS App | No. Auto populated | Yes, requires plus plan. Homepage: sign in on mobile with a paid account. Store listing OS split was not opened this round; Google/Apple both implied by “mobile app”. |
Android App | No. Auto populated | Unknown. A mobile app is documented; Android vs iOS split was not confirmed on a store page. |
Desktop App | No. Auto populated | Yes, requires free plan. Downloadable IDE for the OS; Open VSX + VS Code settings import. |
Chrome Extension | No. Auto populated | No. Not distributed as a Chrome extension. |
Privacy — Bolt New Wins
| Feature | Bolt New | Kiro |
|---|---|---|
GDPR Compliant | Yes. Auto populated | Unknown. Needs an AWS DPA read for Kiro specifically. |
Opt Out of Data Training | Yes. Auto populated | Unknown. Enterprise privacy is advertised; a consumer training opt-out toggle was not checked. |
Enterprise Privacy Mode | Yes. Auto populated | Yes, requires enterprise plan. IAM, SSO, usage dashboards, IP indemnity, GovCloud, centralized billing. |
Workspaces — Kiro Wins
| Feature | Bolt New | Kiro |
|---|---|---|
Projects / Folders | No. Auto populated | Yes, requires free plan. Workspace .kiro/specs and steering; Crew is a persistent local workspace. |