Carina; self-hosted AI agent for operators

Self-hosted AI agent

An agent that finishes tasks on your stack

Carina takes a job, picks tools, and delivers a result; you run it on your infrastructure, not ours.

Running in production? Labyrinth Scout is the separate monitoring and audit layer.

Install Carina Read the docs Carina Cloud
carina
$ npm install -g carina-agent
Setup complete. Agent ready.
You: Summarise last week's support tickets and draft a reply template
Carina: Reading inbox export, grouping themes, writing draft...
Carina: Done. Summary and template saved to ./outbox/
Runs on your machine Real tool execution Remembers past work MIT licensed
How it works

You give a task.
Carina does the work.

Not a chat wrapper. Carina plans steps, calls tools, and reports back. Free to self-host; MIT licensed.

Takes jobs to completion

Research, file edits, shell commands, browser steps, email. Carina acts on your stack instead of handing you a paragraph to copy.

Remembers context

Past conversations and outcomes stay available across sessions so repeat work does not start from zero.

Runs where you host it

CLI, web UI, Telegram, WhatsApp, Slack, or API. Same agent, same memory; you choose the channel.

Full platform details
By audience

Start from your use case

Carina Core is general-purpose. These pages show who each setup is for, what to configure, and one example workflow.

Property

UK deals, HMO checks, property-uk pack and RAG.

Support and ops

Triage tickets, draft replies, save files for review.

Research

Web and arXiv briefs with research-basics.

Deployment options

Run Carina in the browser, on your own server, or both.

Compliance

legal-uk skills on Carina; audit exports on Scout.

All use cases

Hub page plus links to 16 optional skill packs.
Get started

Install in two commands

Self-host with Node.js 22+, or use Carina Cloud if you want a hosted workspace.

npm install -g carina-agent
carina setup
Open Carina Cloud Documentation npm package
Pricing

Carina Core is free

Self-host the full agent at no cost. Monitoring, audit, and kill-switch controls live in Labyrinth Scout, a separate product.

Carina Core

Free
  • Full agent; tools, memory, channels
  • Self-hosted on your infrastructure
  • Optional domain skill packs
  • MIT-licensed source on GitHub
npm install -g carina-agent

Production monitoring, compliance exports, and enterprise plans: labyrinthscout.com

Platform details

Full capability list

Verified counts as of 12 June 2026. Shipped vs configured vs optional: canonical capability inventory.

66+ tools

Base registry; toolsets enable subsets. Web search, files, shell, HTTP, email, code, browser, MCP, and more.

51 LLM providers

Catalog entries; chat uses configured keys or local backends. Switch mid-session with /provider.

19 channel adapters

CLI, web UI, Telegram, WhatsApp, Slack, Discord, email, API server, and more when configured.

Optional

16 domain packs

96 curated skills across engineering, finance, legal, devops, property, and more.

Optional skill writing

Carina can draft new skills after solving complex problems when auto-writing is enabled.

7,500+ tests

Vitest coverage across the agent core. See the public repo for CI status.

Browse skill packs
Carina capability Included in Core
Autonomous task execution
Persistent memory with semantic search
Optional skill writingopt-in
16 skill packs (96 curated skills)optional
19 channel adapterswhen configured
66+ tool execution surfacetoolset-dependent
Switch LLM mid-session (51 catalog; configured subset)when configured
npm package (carina-agent)v0.1.1
MIT-licensed sourceGitHub

Production monitoring and response live in Labyrinth Scout. Capability counts reflect shipped definitions; availability still depends on configuration. See the canonical capability inventory and labyrinthscout.com.