THE HARNESS

Everything an agent needs
to do real work

CRHQ is the harness around the model, so you can build agentic systems without running one: skills, memory, delegation, a real workspace and multi-model routing, on a server that is yours. It ships with system agents and core builders, and you build the rest.

ARCHITECTURE

Hub and Server Model

cr fleet status
CRHQ Hub ● online crhq.ai
├── server-01 ● healthy research-team.crhq.ai
├── server-02 ● healthy dev-team.crhq.ai
├── server-03 ● healthy content-team.crhq.ai
└── server-04 ● deploying ops-team.crhq.ai

4 servers | 12 agents | 128 skills loaded

Each server is a fully isolated VPS: its own server, database, file system, and browser. Agents on one server can't see or touch another. The Hub manages everything from one dashboard.

SKILLS SYSTEM

Capabilities, not prompts

Structured Skills

Each skill is a versioned, distributable package, not a prompt. Skills contain instructions, scripts, API references, and workflows that agents follow precisely.

Create Your Own

Build custom skills in minutes. Database queries, browser automation, API integrations, deployment scripts: wrap any workflow into a reusable skill.

Push to Fleet

Update a skill once, push it to your entire fleet. Every server gets the latest version. No manual updates, no version drift.

cr skills list --active
web-research v3.2 Crawl, extract, analyze web content
code-review v2.1 Review PRs, suggest improvements
seo-optimizer v1.8 Analyze content for search ranking
database-query v4.0 Read-only SQL across project DBs
sentry-monitor v2.5 Query errors, trace stack traces
slack-notify v1.3 Post updates to Slack channels
browser-auto v3.0 Navigate, fill forms, screenshot
deploy-server v2.2 SSH, configure Nginx, manage PM2
... and 120 more
AGENTS

Real tools. Not simulations.

Full Servers

Every agent runs on a dedicated VPS with root access. File system, databases, package managers: the full Linux environment.

Browser Access

Headless Chromium for web scraping, form filling, screenshots, and testing. Agents navigate the web like humans.

Databases

PostgreSQL per server. Agents store data, query results, and maintain structured knowledge across sessions.

Persistent Memory

Agents remember everything. Session context, project knowledge, and cross-agent shared memory that persists across tasks.

Scheduled Jobs

Cron-like scheduling for recurring tasks. Daily reports, weekly audits, hourly monitoring: agents work on your clock.

Credential Vault

Encrypted storage for API keys, tokens, and passwords. Agents access credentials without ever exposing them in output.

Start deploying agents.
Get Started → Schedule a Demo
$299/month per server. Bring your own AI model. Deploy in 90 seconds.