Business Guide
Ship faster with AI developers that work 24/7.
What corvid-agent does for your business
Review every PR
Catches bugs, security issues, and style problems in minutes instead of days — before a human looks at it.
Answer support questions
Responds in Discord or Slack automatically, reducing ticket volume and response time for your team.
Write & fix code
From bug fixes to new features, validated with tests before it opens a PR for human review.
Keep docs in sync
Detects when documentation drifts from code and opens PRs to fix them automatically.
Triage issues
Labels, prioritizes, and responds to new GitHub issues automatically so nothing falls through the cracks.
Generate reports
Weekly summaries of what shipped, what’s in review, and what’s blocked — delivered to your channel.
Your code stays private
corvid-agent runs on your infrastructure. The AI model runs locally (free with Ollama) or via Claude API (~$3/hour of active use). Nothing leaves your network unless you choose it.
Getting started for your team
Step 1: Install (2 minutes)
One person runs the installer on a shared server or their machine:
$ curl -fsSL https://raw.githubusercontent.com/CorvidLabs/corvid-agent/main/scripts/install.sh | bash
✓ Dashboard opens at http://localhost:3000Step 2: Connect GitHub (2 minutes)
# Create a token at github.com/settings/tokens/new with repo scope
$ echo 'GH_TOKEN=ghp_your_token_here' >> .env
$ bun run server
✓ Agent can now read repos, review PRs, and open pull requestsStep 3: Set up your first automation (5 minutes)
Go to Schedules in the dashboard and create a PR review bot:
Name: PR Review Bot
Schedule: */15 * * * * # every 15 minutes
Prompt: Check for open PRs on my-org/my-repo that haven't
been reviewed. For each one, read the diff, check
for bugs and security issues, and post a review.
Approval: Auto
✓ Every PR now gets reviewed automaticallyStep 4: Connect chat (optional, 5 minutes)
# Discord
DISCORD_BOT_TOKEN=your-bot-token
DISCORD_CHANNEL_ID=your-channel-id
# Slack
SLACK_BOT_TOKEN=xoxb-your-bot-token
SLACK_CHANNEL_ID=your-channel-idCommon team setups
Small team (2–10)
- ✓ 1 agent on a team server
- ✓ Auto PR review on main repos
- ✓ Discord/Slack for questions
- ✓ Manual approval for agent PRs
Growing team (10–50)
- ✓ Dedicated server
- ✓ Specialist agents (reviewer, tester, support)
- ✓ Daily velocity & backlog reports
- ✓ GitHub webhooks for real-time triage
Large team (50+)
- ✓ Docker deployment
- ✓ Multi-tenant mode
- ✓ API key per team
- ✓ Supervised approval queues
See Enterprise Guide for details.
What it costs
Ollama (local)
Free AI cost
Runs on your existing hardware. 8 GB RAM minimum. No API calls, no metered billing.
Claude API
~$3/hour of active use
Pay only when the agent is actively working. Idle time costs nothing.
Claude Code CLI
Subscription
Uses your existing Claude subscription. No additional infrastructure cost.
Zero idle cost
The agent only uses AI when it’s actively working — reviewing a PR, writing code, answering a question. When it’s waiting for the next scheduled task, it costs nothing.
ROI examples
PR review automation
Before: PRs wait 1–3 days. Bugs slip through. Senior engineers spend ~2 hours/day reviewing.
After: Every PR reviewed in minutes. Engineers review agent-flagged items only.
Save ~30 hours/month (team of 10)
Support automation
Before: 3 engineers rotate on support. 60% of questions are answered by linking to docs.
After: Agent handles the 60% automatically. Engineers handle complex escalations only.
Save ~40 hours/month
Test generation
Before: Test coverage at 60%. Nobody has time to write tests for existing code.
After: Agent writes tests for untested files weekly, opening PRs for review.
+5–10% coverage/month with zero effort
Security & privacy
Your code stays on your machine
corvid-agent runs locally, not in the cloud. Nothing leaves your network unless you configure external integrations.
API key authentication
Required when the server is exposed beyond localhost. Timing-safe comparison prevents side-channel attacks.
Protected files
.env, credentials, and security-sensitive files are blocked from agent access via basename, path, and symlink checks.
Rate limiting & audit logging
All endpoints are rate-limited. Every action is logged in an immutable audit trail for compliance and review.
For enterprise-grade security (RBAC, multi-tenant isolation, container sandboxing), see the Enterprise Guide or the full Security docs.