Neural Inverse is Open Source →
Docs

Neural Inverse Overview

Neural Inverse is an open-source AI-native IDE built for regulated software development. It combines agentic coding, firmware-aware tooling, and a structured legacy modernization platform into a single governed development environment. Apache 2.0. No credit card required.

Install

curl -fsSL https://cdn.neuralinverse.io/ce/install.sh | bash

Windows:

irm https://neuralinverse.com/win | iex

Core Features (Open Source)

Power Mode

Power Mode is Neural Inverse's agentic coding engine. It spawns three specialized sub-agents that work in parallel:

  • Explorer — read-only research agent. Scans the codebase, gathers context, maps dependencies. No write access.
  • Editor — makes targeted code changes scoped to specific files approved by the parent task.
  • Verifier — runs commands, executes tests, reads lint errors, and reports results.

Each sub-agent has an isolated tool scope. The Explorer cannot write files. The Editor cannot run arbitrary terminal commands. The Verifier closes the loop by confirming changes work.

Power Mode sessions track every message, tool call, and code change. Each session has a status (idle, busy, error, compact) and a summary of additions, deletions, and files changed.

Firmware Development

Firmware Development is a purpose-built environment for embedded and safety-critical hardware work:

  • MCU database with 357+ variants across STMicroelectronics, Nordic, Espressif, NXP, TI, Microchip, Renesas, Infineon AURIX, and more
  • Supported cores: Cortex-M0 through M85, Cortex-A/R, RISC-V, Xtensa (ESP32), TriCore, AVR, PIC, MSP430, and others
  • SVD auto-load — parses .svd files and injects register maps, bit fields, and reset values into the LLM context
  • Datasheet intelligence — extracts timing constraints, errata, and peripheral configs from PDF datasheets using your LLM
  • Hardware agent tools — 22+ tools covering peripheral queries, serial monitor, build analysis, code generation, simulation discovery, and compliance checking
  • Compliance frameworks — MISRA-C 2012/2023, CERT-C, IEC 62304, ISO 26262, DO-178C, IEC 61508, AUTOSAR
  • Session persistence — MCU config, loaded SVD files, datasheets, and active peripheral survive IDE restarts via Firmware.inverse

Legacy Modernization

Legacy Modernization is a structured 5-stage migration platform:

StageWhat happens
1. DiscoveryScans source codebase, detects languages, maps dependencies, identifies compliance patterns
2. PlanningGenerates migration plan, CPM critical-path scheduling, defines migration units
3. MigrationFile-by-file translation using your LLM, 36 supported language pair profiles
4. ValidationFingerprint comparison, compliance checks, unit review
5. CutoverReadiness gate, audit bundle export

Supported source languages include COBOL, FORTRAN, RPG, PL/SQL, Natural, Assembler, AUTOSAR ARXML, IEC 61131-3 (PLC ladder/structured text), and more. Each session is paired via a Modernisation.inverse project file written to both source and target roots.

Supported migration patterns include bare-metal-to-RTOS, PLC-to-IPC, SCADA modernization, OT/IT convergence, OPC-UA migration, and general language pair translations.

BYOLLM

Neural Inverse never routes your code through Neural Inverse servers. You bring your own LLM API keys — they stay on your machine.

Supported providers: Anthropic Claude, OpenAI GPT-4, Google Gemini, Ollama (local), AWS Bedrock, and any OpenAI-compatible endpoint.

See LLM Providers for setup.

Where to Start

Open Source

Neural Inverse is fully open source under Apache 2.0.


Was this page helpful?