InsForge Launch
The backend that works automatically with AI coding agents.
Published November 18th, 2025
Today we are officially launching InsForge.
We spent the past four months building InsForge and testing it with a wide range of early developers. Companies like Zeabur and Peak Mojo have already used InsForge to move faster and launch real features during this period.
With their feedback and the momentum from our early community, we are now ready to open InsForge to everyone.
Early Traction:
Databases Launched
2,079
GitHub Stars
865
Pull Requests Resolved
470
Forks of InsForge Repositories
120
01 Performance
We started InsForge with the belief that natural language will become the new way to build software. A prompt driven workflow only works when coding agents understand the backend correctly. Without accurate context they hallucinate, apply incorrect changes, and fall into long debugging cycles.
We designed our entire MCP stack for agent oriented development. Every endpoint is built around context engineering, returning clean and reliable information from a single source of truth. We wrote more about this idea in our previous blog, Why Context is Everything in AI Coding.
We ran MCPMark Benchmark (an open source benchmark) to evaluate MCP servers, running 21 tasks across 6 Postgres databases. These datasets range from small relational schemas to large enterprise scale systems. The tasks cover analytical reporting, data migration, CRUD logic, query optimization, RLS policies, triggers, and vector search.
Speed
Lower = better
1.6x
Faster
InsForge
150s
Supabase
239s
Postgres
215s
Efficiency
tokens / run
30%
Fewer Tokens
InsForge
8.2M
Supabase
11.6M
Postgres
10.4M
Accuracy
1.7x
Higher Accuracy
InsForge
47.6%
Supabase
28.6%
Postgres
38.1%
Benchmarks were run and evaluated using the same coding agent and model stack, Claude Code with Sonnet 4.5, to ensure consistent and reliable comparisons across all tasks and datasets. Each test case was executed 4 times to reduce variability. Accuracy is reported using Pass⁴ (strict): a task counts as successful only if the model passes all 4 runs.
Higher accuracy and lower token usage produce a smoother agent workflow. Agents make fewer mistakes, require fewer retries, and complete multi step backend updates with less time and cost.
For developers this means a faster and more predictable prompt driven development experience. InsForge strengthens this workflow by providing agents with clean context, stable tools, and a backend that is designed from the ground up for natural language development.
02 Security
As an infrastructure provider built for AI centered development, we set strict security standards from the beginning. Authentication data is encrypted, and database access is enforced with anonymous rules and RLS based policies to maintain proper isolation.
Each service is designed with clear boundaries. Sensitive operations are validated inside the backend rather than trusting client input. Secrets and configuration values are never exposed to coding agents, storage uses presigned URLs for controlled file access, and functions run with a managed secret store.
Every new feature goes through an internal security review with automated checks and a dedicated cycle of manual testing. We pay special attention to how coding agents interact with the system, making sure their workflows follow safe patterns and apply secure configurations by default.
03 Features
Here's the quick overview for the backend services InsForge provides, you can use the whole platform or just the services you want.
MCP
The InsForge MCP server gives coding agents access to the full backend through a single, structured interface. Agents can read project metadata, inspect database schemas, update backend configuration, manage storage, and deploy functions through MCP tools. InsForge works perfectly with agents:
And More...
Authentication
InsForge provides a complete authentication service so you can add sign-in features with minimal setup. It supports email and password, social sign-ins, and both code and link based verification. InsForge also includes ready to use React components for sign-in and sign-up flows, so you can add authentication UI without writing front-end logic.
Database API
InsForge provides a fully managed Postgres database that you can use immediately without setting up an ORM. With the InsForge SDK you can perform CRUD operations directly and work with your data in a simple and consistent way.
Table View
InsForge dashboard provides GUI for your database tables, where you can easily manage your data or import CSV.
SQL Editor
You can also run queries directly from the SQL editor in the InsForge dashboard without installing external tools.
Storage
Store and manage files directly through the InsForge SDK using secure URLs and database controlled access. From the dashboard you can view files, create buckets, and switch between private and public visibility.
Functions
Run backend logic without managing any infrastructure. Functions work seamlessly with our Secret Manager, letting you store and use API keys securely inside your application.
AI Integration
Use AI models from OpenAI, Anthropic, Gemini, and Grok through a single unified SDK. Everything works out of the box with no API key setup or credit management. Usage is billed directly through InsForge with simple usage based pricing and 0% fees.
04 Open Source
Open source has been part of InsForge from day one. Our platform is built on PostgreSQL and benefits directly from the strength and reliability of the open source ecosystem.
In October we participated in Hacktoberfest and received contributions from thirty one developers, along with more than two hundred new stars on our repositories. Their feedback and pull requests helped us refine the product and improve areas that matter to real users.
The open source community has played a major role in our progress, and we will continue investing in it as we grow. You can explore our GitHub here and follow our public roadmap to see what we are building next.
05 Partnerships
Supporting early teams has been part of our mission from the beginning. Many founders want to build with AI but face limits around compute, tooling, or backend setup. We want to remove those barriers.
We have partnered with programs such as StartX Accelerator at Stanford and Grid110 to provide their members with expanded InsForge compute resources and premium support. This helps early teams iterate faster and focus on building real products instead of managing infrastructure.
We remain committed to supporting early founders and will continue expanding these partnerships.
06 Thank You to Early Users!
To thank everyone who supported us from the beginning, all current Pro users will receive one full year of Pro subscription at no additional cost. You do not need to take any action. The upgrade has already been applied to your account.
07 What's Next
Our pace of shipping never slows down. We continue to listen to our community and release new features almost every day. Looking ahead, our next major milestone will be InsForge Launch 2 in February 2026, and here are some of the areas we will be focusing on.
Infrastructure
- Multi Region: Rolling out new regions to improve performance for users across more than 100 countries.
- Multi Cloud: Deploying InsForge across multiple datacenter providers for higher resilience.
- Server Status Reporting: Introducing a public status page for real-time service visibility.
AI Assisted Development
- Backend Branching and Versioning: Staging and sandbox environments so AI changes remain safe with human approval.
- AI Backend Advisor: AI powered insights that detect issues and suggest improvements for backend structure.
- Project Memory: Stores key information about your project and backend so coding agents always understand the current structure, past decisions, and recent changes. This allows agents to continue work instantly without re-explaining context or risking inconsistent updates.
Developer Tools
- More Client Libraries: We are expanding our client libraries to support more languages and frameworks, especially for app developers building with mobile and cross-platform stacks. The community has already helped us grow our SDK support, and we welcome contributions from anyone who wants to bring InsForge to their favorite language or framework.
- CLI: A command line interface for working with InsForge from your local environment.
We are just getting started, and as we prepare for our next major milestone we would love your feedback. Email us at info@insforge.dev or join our Discord community and tell us how we can help you build faster.
Start Building Real Apps
