Connect Codex, Claude Code, Cursor, or any MCP/HTTP-capable agent to one tenant-scoped system for domains, mailboxes, warmup, campaigns, and replies. Your agent owns strategy and copy; Coldrig runs the durable machinery.
Launch founder outreach for Northstar. Keep it personal and protect the brand.
I’ll set up isolated infrastructure, warm each mailbox, then launch only when the rig is healthy.
setup_infrastructure
8 mailboxes
infrastructure_status
day 12 of 14
launch_campaign
waiting for health
The agent stays the brain.
Coldrig handles the machinery.
You stay in control.
Most outreach stacks make your agent juggle registrars, mailbox vendors, sequencers, warmup tools, and inboxes. Coldrig collapses the machinery into a few high-level intents.
Your agent asks for outcomes, not vendor plumbing. Twenty-eight carefully scoped tools keep context small and execution legible.
Explore the tool surface →Dedicated domains and mailboxes keep reputation contained. Nothing is shared across customers.
Send caps, suppression, unsubscribe handling, and health-based pausing are enforced below the agent layer.
Your agent operates the system. You get a calm control room for health, activity, campaigns, and conversations.
Coldrig is built around the way agent-led work actually happens: the human sets the outcome, the agent plans and operates, and the product makes every action visible.
ColdrigPaused Atlas / inbox 04bounce rate crossed 3%
Launched Founder signal420 leads · 3 steps
Needs your review8 positive replies
Product illustration · simulated account data
Start with a fault-injecting sandbox that simulates provisioning, warmup, replies, bounces, timeouts, and rate limits — a safe technical pilot before you commit. Real sending is live in production; signup, billing, screening, and real mailbox provisioning are fully self-serve and automatic, and mailbox send-authorization completes on our side.
$ npx agent-cold-email demo
Creating isolated demo tenant…
✓ infrastructure ready
✓ replies + failures simulated
↳ no real spend · no real sends
_
Plain-English status: the API, MCP endpoint, npm CLI, sandbox, and real sending are all live. Signup, billing, screening, and real mailbox provisioning are fully self-serve and automatic; the only remaining step is mailbox send-authorization completing on our side. Coldrig never guarantees inbox placement. Read the honest FAQ →
agent-cold-email is built for a different case: one platform, one bearer token, and 28 curated tools that already cover both the infrastructure layer and the sequencing/reply layer a DIY vendor stack or a sending platform leaves to a separate vendor.
| Dimension | DIY vendor stack | Sending platform + API | Coldrig |
|---|---|---|---|
| Domain + mailbox provisioning | Generate candidates, buy each, configure SPF/DKIM/DMARC/rDNS per domain, verify manually. | Most often paired with a separate domain/mailbox infrastructure vendor (Infraforge, Mailforge, PrimeForge, or a bulk-mailbox provider like Mailscale/InboxKit) — the sending platform and the infrastructure layer are treated as two different purchases. | One setup_infrastructure call; DNS is verified automatically before a mailbox is provisioned on it. |
| Warmup state | Separate vendor account, manual per-mailbox warmup tracking across weeks of sessions. | Tracked and re-derived across two separate systems, across sessions. | Provisioned and warmed as part of the same call; infrastructure_status reports current state on demand — no state to re-derive. |
| Campaign execution | Self-built or separate tool: scheduler, caps, bounce/reply/unsubscribe detection, threading. | Campaigns, leads, email accounts, warmup, smart delivery, smart senders, webhooks, sequences, analytics. | launch_campaign, campaign listing, pausing, and results. The caller supplies leads and copy. |
| Replies + lead disposition | Bounce/reply/unsubscribe detection, threading — self-built or a separate tool. | 6 categories including reply classification, unsubscribe handling, lead routing. | Unified inbox, full thread, reply, read state, and labels. |
| Guardrails enforced server-side | Implemented by you or your agent, correctly, every campaign, with no server-side backstop if a step is skipped. | One cautionary account found during research described buying 30 mailbox accounts through an infrastructure vendor — all of which went to spam. | Enforced server-side by default (suppression, send caps, complaint auto-pause, footer address injection) — not dependent on the agent remembering. |
| One credential / one stateful surface | Registrar API key + mailbox-vendor API key + (optionally) sequencer credentials — three-plus separate secrets, three-plus auth models. | Two separate API keys/logins, two separate auth models. | One bearer token, one auth model, resolved fresh per call; usage, quota, warmup state, and billing live durably in the platform. |
| Agent-facing interfaces | Every new agent session touching this project re-derives vendor state from three-plus systems. | 31-38 reported tools via an official first-party MCP — but even a fully MCP-driven setup still needs one manual login to mint an API key. | Hosted streamable-HTTP MCP, REST/OpenAPI, and the published npm CLI. |
| What it does not claim | No deliverability, open-rate, or reply-rate comparison is made anywhere on this page. | We have not independently verified these numbers, and we do not have a comparable production track record of our own yet. | Coldrig never guarantees inbox placement. |
Sourced from the DIY comparison, Smartlead vs Instantly vs agent-cold-email, and infrastructure vs. sending platform — full sourcing and caveats on each page.
Run the free sandbox now, or go live in minutes: call POST /checkout for a hosted Stripe payment link and pay — signup, billing, screening, and real mailbox provisioning are fully self-serve and automatic, with mailbox send-authorization completing on our side. Want product updates instead of going live right now? Drop your email below.
By joining, you agree to the privacy policy. Unsubscribe anytime.