Neural Inverse Build
Documentation

Version Latest

Editor

The Neural Inverse Editor, a desktop development environment with AI assistance for regulated engineering teams, what it does today and where each task is documented.

The Editor (named Neural Inverse on your system) is a desktop development environment for hands-on work with AI assistance: chat, agent sessions that edit your code, and edits you review before you keep them. It is built on Code - OSS, the open-source core of Visual Studio Code, so the editing, debugging, terminal, source control and extension features of Code - OSS are there. Neural Inverse adds sign-in to your organisation, organisation policy, approvals with electronic signatures, a tamper-evident activity record, evidence packs, Checks, organisation context and bring-your-own models.

The Editor is installed together with the Agents app. Each app runs on its own, with its own process, icon, data folder (.neural-inverse for the Editor, .neural-inverse-agents for the Agents app) and sign-in session.

Maturity

The Editor is pre-release. Interfaces and data formats may change between versions. Several features talk to your organisation's Neural Inverse Govern deployment (powered by Keystone), and some of them have only been tested against servers written from the published contract. Each page says what is verified and what is not.

What you can do

TaskPage
Sign in, see which server and account you use, sign outFirst launch and signing in
Open folders, trust a workspace, see which project a folder belongs toWorkspaces, trust and project binding
Ask questions in chat, run agent sessions, review and keep editsChat and agent sessions
Connect model providers, local models and your organisation's modelsModels
See compliance findings and obligations as you workChecks
Understand what your organisation's policy enforces on this deviceOrganisation policy
Request and sign approvalsApprovals and signatures
Set your intent, link work items, trace requirements, baseline releasesIntent, projects and releases
See and export what was recorded on this deviceActivity Log and audit upload
Build and verify evidence packsEvidence packs
Add your organisation's code, documents and issues to prompts, share contextOrganisation context
See when another team's agents work on the same codeTeam work
Keep credentials out of agent terminals, run a program with vault secretsSecrets and agent terminals
Look up a setting, a managed policy or a commandSettings and policies, Commands and keys
Fix a problemTroubleshooting

For firmware, FPGA and HDL, PLC, DCS, mainframe, energy and automotive work, see the domain module sections, for example Embedded.

How the Editor fits with the rest of Neural Inverse

  • Sign-in. You sign in to your organisation's Govern deployment with your own account. While you are signed out, a sign-in screen covers the window. See Get started.
  • Agent host. Agent sessions and model calls run in a separate agent host process. On macOS one agent host is shared by the Editor and the Agents app and keeps running when either app quits. On Windows and Linux each app runs its own host unless the environment variable NI_SHARED_AGENT_HOST=1 is set, because the shared host's process behaviour has not yet been validated there.
  • Policy. Your organisation's policy arrives signed and is verified on the device. A modified build can remove any check inside the Editor, so what must hold is enforced by Govern (its model gateway, for example) and, for merges and releases, by your CI. See Organisation policy and Govern.
  • Audit. By default nothing in the activity record leaves the device. Your organisation can turn on uploading of agent events, and separately a person's own actions, to its own Govern deployment. No telemetry is sent to Neural Inverse.

Not in this release

  • Legacy modernisation. The source contains a modernisation workflow, but no app loads it today: a test fails if an app loads it without its governance contribution. It is not documented until it ships.
  • The older agent manager, power mode and Ollama auto-setup. Their code is in the repository but neither app loads it. Local models are connected on the Models page instead.
  • Sharing running agent sessions between the Editor and the Agents app is in progress. See Working with the Editor for what is shared today.

Licence

The Editor is proprietary and licensed under your organisation's agreement with Neural Inverse. Code inherited from Code - OSS stays under the MIT License. See licence.

Last reviewed 17 October 2026View as Markdown

Editor · Docs · Neural Inverse