Key takeaways
- Niteshift delivers merge-ready PRs with end-to-end browser verification — screenshots prove the code works before review
- Supports Claude Opus 4.6 and Codex 5.3 with customizable cloud dev environments
- Remote MCP servers (Linear, Notion, Figma) via OAuth enable deep tool integration
- Prototype-to-production pipeline converts Lovable or Figma designs into working features with real API calls
- Currently in waitlist/pre-launch with active weekly development cadence
FAQ
What is Niteshift?
Niteshift is a background coding agent that takes prompts, bug reports, or prototypes and returns merge-ready pull requests with browser screenshots proving the changes work.
How much does Niteshift cost?
Niteshift is currently in waitlist/pre-launch with no public pricing. You can join the waitlist at niteshift.dev/waitlist.
What models does Niteshift support?
Niteshift supports Claude Opus 4.6 (default) and Codex 5.3 (via ChatGPT subscription authentication).
Company Overview
| Quick Reference | |
|---|---|
| Website | niteshift.dev |
| Founded | ~2025 |
| Headquarters | New York, NY |
| Founder | Israel Asefa |
| Funding | Not disclosed |
| Status | Waitlist / Pre-launch |
Niteshift is a background coding agent platform that "closes the loop" — you give it a prompt, bug report, or prototype and get back a merge-ready PR with screenshots proving it works. Founded by Israel Asefa, a senior backend engineer with experience in Java and Spring Boot production systems, Niteshift operates from New York, NY.
The platform sits in the same "autonomous coding agent" category as Devin (Cognition) and Tembo, but differentiates through end-to-end browser verification and a prototype-to-production pipeline that converts design tools into working features.
Product Analysis
How It Works
Niteshift runs full dev environments in the cloud with customizable agentic setup. When you assign a task — via chat, Slack, or slash commands — the agent writes code, then verifies it by opening a real browser to test frontend, API, and database changes. It iterates until everything works, then opens a PR with screenshots as proof.
Key Capabilities
End-to-End Verification Unlike agents that only generate code, Niteshift opens a real browser to test changes, capturing screenshots and API responses to prove functionality before submitting a PR. Verification modes include browser automation, CI fixes, and screenshot posting.
Autofix CI & Code Review Niteshift automatically responds to all review comments and fixes all CI failures in every PR it opens. This reduces the back-and-forth loop that plagues other coding agents.
Prototype to Production
The /lovable command points Niteshift at a Lovable app prototype, and it builds the feature into production — frontend and backend — replacing mocked data with real API calls and database reads, then doing multiple polish passes to match the original design. Figma prototypes are also supported via MCP integration.
Model Support Niteshift supports Claude Opus 4.6 (default) and Codex 5.3 (via ChatGPT subscription authentication). Both Claude Code and Codex CLIs are available in the terminal for all tasks regardless of which model runs the primary task.
MCP Server Support Remote MCP servers for Linear, Notion, and Figma connect via built-in OAuth, extending the agent's capabilities into project management and design tools.
Cloud Dev Environments Full dev environments run in the cloud with agentic setup scripts. A "Manage" interface lets users iterate on setup scripts with automatic restarts.
Static IP Egress Proxy For teams with IP-restricted databases or firewalled APIs, Niteshift provides a dedicated static IP for allowlisting.
Slack Integration Channel mentions pass relevant conversation context to the agent, including attached files and images.
Development Velocity
Niteshift maintains a weekly changelog with consistent feature releases:
| Date | Highlights |
|---|---|
| Feb 16, 2026 | Remote MCP servers, file references in chat, improved Slack integration |
| Feb 9, 2026 | Claude Opus 4.6, Codex 5.3, static IP egress, task sidebar |
| Feb 2, 2026 | Background terminals, verification mode selection |
| Jan 26, 2026 | Lovable prototype integration, image paste in terminal, browser session persistence |
This cadence suggests an actively developed product with rapid iteration, though the product remains in pre-launch waitlist status.
Strengths
- Verification-first approach — Browser screenshots and API capture provide proof of working code, not just generated code
- Rapid iteration cycle — Weekly changelog shows consistent, meaningful feature development
- Prototype pipeline — Lovable/Figma to production is a unique workflow few competitors offer
- MCP ecosystem — Remote MCP servers via OAuth enable extensible tool integrations
- CI autofix — Automatic CI failure resolution reduces PR churn
Cautions
- Pre-launch risk — Waitlist-only with no public pricing; product viability unproven at scale
- Unknown funding — No disclosed backing raises sustainability questions
- Limited visibility — No public GitHub repo, minimal external coverage or community traction
- Small team — Founder-stage company competing against well-funded players like Devin ($175M+) and orchestrators like Tembo ($20M)
- Model dependency — Relies on Claude and Codex without custom model optimization (unlike Devin's SWE-1.5)
Competitive Positioning
Niteshift occupies the "background autonomous agent" segment alongside Devin and Tembo, but with a distinct focus on verification and prototype conversion:
| Dimension | Niteshift | Devin | Tembo |
|---|---|---|---|
| Verification | Browser + screenshots | Basic | Via agent |
| Model flexibility | Claude, Codex | Custom only | 5+ agents |
| Prototype pipeline | Lovable, Figma | — | — |
| MCP support | Yes (OAuth) | — | Yes |
| Automations | — | Events only | Cron + Events |
| Integrations | Slack, Linear, Figma | Slack, GitHub | 11 native |
| Pricing | TBA (waitlist) | ~$500/seat/mo | $60-200/mo |
| Status | Pre-launch | GA | GA |
Bottom Line
Niteshift brings a compelling verification-first approach to autonomous coding agents — the idea that a PR should come with proof it works, not just code diffs. The prototype-to-production pipeline and MCP integration show thoughtful product design. However, as a pre-launch, unfunded (publicly) startup competing against well-capitalized players, execution risk is high. Worth watching if browser-verified PRs and prototype conversion are priorities for your workflow.
Research by Ry Walker Research • methodology
Disclosure: Ry Walker is CEO of Tembo, which competes in this category.