InsForge V2
The backend built for agentic development.
Published March 9th, 2026

Today we are launching InsForge 2.0
Since our last launch in November 2025, we've grown fast. The number of databases created on InsForge increased by 500%, and almost 99% of all operations and requests now come from agents instead of humans. Provisioning, configuration, migrations, and runtime actions are executed by agents, not through dashboards, CLIs, or hand written scripts. This shift made one thing clear: agent experience is the new developer experience.
Agentic development requires agent native infrastructure, which is what we have been building over the past few months. InsForge 2.0 is the result.

01 Performance
We reran the same MCPMark database benchmark, same 21 Postgres tasks, same methodology, now on Claude Sonnet 4.6. In benchmark result v2, InsForge improves its agent native performance and the gap with traditional backend MCPs widens.
Accuracy (pass@4)
+14% higher accuracy than Supabase
Feb 2026
InsForge
76.2%
↑ 23.1% since Nov 2025
Supabase
66.6%
↓ 6.8% since Nov 2025
Previous benchmark (Nov 2025)
InsForge
61.9%
Supabase
71.4%
Tokens per Run
59% fewer tokens than Supabase
Feb 2026
InsForge
7.3M
↓ 11.0% since Nov 2025
Supabase
17.9M
↑ 54.3% since Nov 2025
Previous benchmark (Nov 2025)
InsForge
8.2M
Supabase
11.6M
The newer SOTA model reasons more extensively when backend context is incomplete. Supabase MCP server therefore triggers more discovery queries, more verification steps, and significantly higher token usage.
With our latest progressive disclosure and semantic layer updates, agents receive the relevant schema, RLS, and relational context upfront. They spend less time exploring and more time executing. This increases accuracy and reduces tokens at the same time.
02 Multi-Region
Our developers and builders are distributed globally, and adding more regions was one of the most requested features. We are starting with four locations where the majority of those requests came from:
- us-east (Virginia)
- us-west (California)
- eu-central (Frankfurt)
- ap-southeast (Singapore)
This brings infrastructure closer to your users, improves latency for applications, and makes InsForge more practical for production workloads in different parts of the world. We will continue to add more regions as our community grows.
03 Reliability
As an infrastructure provider, availability is one of our highest priority goals. We have been working to improve uptime and ensure user data is backed up in a secure and encrypted way.
We are also launching a public status page to track uptime across our services and critical infrastructure: https://insforge.status.io/

04 New Features
Over the past few months we shipped a set of new features across the platform. Here is what we added.
Deployment
InsForge now supports site deployment via the Vercel Platform API, enabling fully automated build and deployment workflows.
Realtime
We introduced a Realtime module built on WebSockets with a pub sub model. You can subscribe to channels for live updates, replay recent messages for state sync, and apply RLS based permissions by default. This makes it straightforward to build chat, multiplayer, collaborative, and other event driven applications on top of InsForge.
Instance Management
Projects can now scale to higher capacity compute instances when workloads demand more CPU, memory, or throughput.
Mobile SDKs
You can build mobile apps on InsForge using our native Swift and Kotlin SDKs. They provide a type safe and consistent way to interact with InsForge backend services from iOS, macOS, Android, and JVM applications.
Remote MCP
We introduced Remote MCP Servers, so you no longer need to run MCP locally over stdio. InsForge hosts your MCP server, secures it with OAuth, and exposes it over the network. Coding agents, Claude Artifacts, ChatGPT connectors, and other MCP clients can connect directly to your project with zero local setup.
CLI
InsForge CLI is the recommended way for agents to use InsForge. It gives agents a clear, scriptable interface for creating projects, updating backend config, and deploying apps.
Create a project with a single command:
Agent Skills
Agent Skills gives coding agents a structured way to use InsForge. They encode common workflows and CLI patterns so agents don't spend tokens discovering MCP tools, making backend operations faster and more reliable.
VS Code Extension
InsForge now provides a VS Code extension to connect and manage projects and MCP servers directly from your editor.
05 Pricing Update
Pricing has been one of the hardest things for us to get right. After talking to early users and looking at real usage, we realized our previous three tiers were too complicated and, in some cases, asked people to pay before they were getting real value. So we are simplifying to two tiers:
- Free: $0/month with 2 fully dedicated instances and unlimited MCP usage
- Pro: $25/month for higher limits, production workloads, and pay as you go usage
You can use the full InsForge stack on Free without a credit card and move to Pro only when you actually need more capacity. If you are already on a paid plan, nothing changes. Your plan becomes a legacy tier with the same pricing and limits, and no migration is required. Pricing should grow with you, not block you from getting started.
Here are our commitments:
- as we grow, we will offer more value for less across all tiers, not the other way around
- pricing will stay simple and predictable
- if something feels unclear, unfair, or limiting, tell us and we will fix it
- if you need credits or technical support to get started, reach out and we will do our best to help
Full details: https://insforge.dev/pricing
06 What's Next
As models and agents continue to improve, we believe agentic development is already becoming the default way software is built. Infrastructure needs to evolve alongside this shift.
Here are two areas we are focusing on next:
1. Infrastructure Reliability
- Global Expansion: Expanding our footprint to bring InsForge closer to developers worldwide.
- Architecture Improvements: Continuously improving the underlying infrastructure to support growing agent driven workloads.
- Availability and Stability: Strengthening reliability so InsForge can support production systems at scale.
2. Agent-Native Backend
- Backend Branching and Versioning: Staging and sandbox environments so agents can safely experiment before changes reach production.
- AI Backend Advisor: AI powered insights that detect issues and suggest improvements for backend structure.
InsForge is built together with our community. Your feedback helps shape what we build next. Email us at info@insforge.dev or join our Discord community and tell us what you would like to see.
The Backend Built for
Agentic Development


