neuralinverse [options] [prompt] Start an interactive session, optionally with a first prompt
neuralinverse -p "prompt" [options] Run once non-interactively and print the result
neuralinverse <command> [args]ni is the same program. neuralinverse --help prints the built-in help.
Commands
| Command | What it does | Page |
|---|---|---|
doctor | Check installation, configuration and connectivity | Install and update |
setup [provider] | Connect a model provider (key checked live, stored in your keychain) and pick defaults | Providers and models |
models [provider] | List models available from configured providers | Providers and models |
config [show|path|validate] | Inspect the effective configuration | Settings reference |
settings export <file>, settings import <file> [--replace], settings sync [--repo <git-url>] | Move settings, key bindings, commands, agents, skills and styles between machines | Settings reference |
sessions | List saved sessions for this folder | Sessions and context |
sessions export <id> --html [--out file] | Write a session as a self-contained HTML page | Sessions and context |
login [--base-url URL] [--device | --browser] [--no-browser] | Sign in to Govern | Sign in to Govern |
logout | Sign out (revokes the session, deletes local tokens) | Sign in to Govern |
whoami [--json] | The signed-in account, device and policy state | Sign in to Govern |
account [show] [--json] | Your account as the Editor and the Agents app show it | Sign in to Govern |
policy [show|refresh] | Show or refresh your organisation's signed policy | Sign in to Govern |
context <status|retrieve|attributions|memory|work|overlap|agent-files|watch> | Govern's Context Platform for this folder | Organisation context |
gov <gate|release|approvals|policy|hooks|trace|hook|trailer> | Organisation governance: gates, approvals, policy, git hooks | Permissions and modes |
checks run [paths] [--staged | --all] [--format text|json|sarif] [--gate agent|commit] [--out file] | Check changes against your organisation's rules | Checks |
checks status | Engine and packs, policy, waivers, last run | Checks |
checks hook install [--force], checks hook uninstall | The pre-commit gate | Checks |
activity [status|log [n]|flush|purge] | What activity capture records and sends | Activity record |
mcp list, mcp add <name> [--user] [KEY=VALUE...] -- <command> [args...], mcp add <name> --url <url> [--sse], mcp remove <name> [--user], mcp login <name>, mcp logout <name> | Manage MCP servers | MCP servers |
plugin [list|install|uninstall|update|enable|disable|info|validate] | Manage plugins | Plugins |
env, env set <none|embedded|hdl> | Show or set the project's developer environment | Firmware and hardware environments |
team "<goal>" [--allow-dirty] [--apply] [--max-tasks n] [--concurrency n] | Split a large task across parallel agents in git worktrees | Goals, team runs and schedules |
schedule [add|list|remove|run|tick|daemon|install] | Run goals on a cron schedule | Goals, team runs and schedules |
serve [--port n] [--host h] [--root dir] ... | Run the HTTP and SSE API | SDK, server and embedding |
web [serve options] [--no-open] | Run the API with the browser interface and open it, signed in | Web interface |
acp | Run as an Agent Client Protocol agent on standard input and output | Editors |
host [--stdio | --socket <path> --token-file <file>] | Serve an embedding host | SDK, server and embedding |
eval [run|report|list|validate] | Measure the agent on the eval tasks | Evals |
neuralinverse context help, checks help, serve help and eval help print the options of those commands; neuralinverse gov without arguments prints its help.
Global options
| Option | Meaning |
|---|---|
-p, --print | Non-interactive mode; prompt from the arguments and/or standard input |
--output-format <f> | text, json or stream-json (with --print; default text) |
-m, --model <ref> | Model as provider/model |
--mode <mode> | ask, auto-edit, read-only, full-auto |
--allow <rule> | Allow rule for this run (repeatable) |
--deny <rule> | Deny rule for this run (repeatable) |
-c, --continue | Continue the most recent session in this folder |
-r, --resume [id] | Resume a session by id or a short prefix of it; a picker if omitted |
--max-turns <n> | Maximum model turns per prompt |
--max-cost <usd> | Stop once this session's estimated cost reaches the amount (exit code 3 with --print) |
--max-time <minutes> | Stop once this session has run this long (exit code 3 with --print) |
--reasoning <level> | off, low, medium, high |
-C, --cwd <dir> | Run as if started in that folder |
--add-dir <dir> | Let tools use an extra folder (repeatable) |
--config <file> | Load an extra settings file (repeatable) |
--mcp-config <file> | Load MCP servers from a JSON file {"mcpServers": {...}} (repeatable) |
--no-mcp | Do not start MCP servers |
--fullscreen | Use the full-screen interface |
--inline | Use the inline interface (the default) |
--append-system-prompt <text> | Extra instructions appended to the system prompt |
--intent <ref> | The work item or change request this session serves (PROJ-123, #12, AB#7, an https link), for policies that require one |
--no-session-persistence | Do not save this session |
--goal <objective> | Work unattended toward an objective (default mode auto-edit) |
--goal-resume <id> | Continue a goal run |
--checkpoint-every <k> | With --goal: commit a snapshot to ni/goal/<id> every k iterations (0 = off) |
--verbose | Progress on standard error in --print mode |
--debug | Diagnostic details on errors |
-v, --version | Print the version |
-h, --help | Show help |
Options take their value as the next argument or after = (--model=openai/gpt-5). -- ends the options.
The built-in help lists login [--issuer URL]; that flag is retired and refused. Use --base-url.
Exit codes
| Command family | Codes |
|---|---|
Sessions and -p | 0 success; 1 error, turn limit or refusal; 2 invalid usage; 3 cost or time limit reached (with --print); 130 interrupted |
account show | 0 signed in; 1 signed out or the session has ended; 2 usage error |
context | 0 done; 1 Govern unreachable or error; 2 usage error; 3 overlaps with --strict-exit; 4 needs a person; 5 not signed in; 6 refused by your organisation. See Organisation context |
checks run | 0 passes; 1 blocked; 2 blocked only because an analyser failed under checks.failClosed |
gov | 0 pass; 1 refused or failed; 2 error or tampering found |