Coldrig vs Salesforge / Forge Stack for AI agents
Choose the Forge Stack today when real sending, LinkedIn, lead sourcing, push webhooks, and an established production product are requirements. Pilot Coldrig when you specifically want your existing coding agent to operate a smaller, single-token, managed infrastructure-to-reply surface—and can wait for real sending. Coldrig is not the production winner today.
Fast decision table
| Requirement | Salesforge / Forge Stack | Coldrig |
|---|---|---|
| Real cold-email sending now | Yes; production product and trial | No; public sandbox only |
| MCP for coding agents | Hosted endpoint for Claude Code, Cursor, and other MCP clients | Hosted streamable-HTTP endpoint plus REST and npm CLI |
| Authentication shape | One endpoint; separate product headers/keys for each Forge product used | One tenant bearer token for all 17 intents |
| Infrastructure | Mailforge, Infraforge, or Primeforge; Warmforge for warmup | One intended managed lifecycle behind setup_infrastructure |
| Lead sourcing | Available through Leadsforge | Not included |
| Email + LinkedIn | Supported | Email only |
| Webhooks | Salesforge's official MCP help lists webhooks in the supported surface | No public push-webhook subscriptions; poll activity |
| Agent surface philosophy | Broad, product-specific controls across the Forge products configured | 17 outcome-level intents across Coldrig's documented lifecycle |
| Production track record | Established product with public customers and reviews | None yet |
Where the Forge Stack is stronger
- It ships real sends now. This is decisive if the user is buying a working production stack rather than evaluating architecture.
- Its agent story is explicit. Salesforge publishes setup instructions and operating workflows for Claude Code and Cursor, not merely a generic API.
- It covers more jobs. Lead search and enrichment, email and LinkedIn, AI-assisted sequence work, webhooks, placement tests, and several infrastructure choices are available across the Forge products.
- It has evidence Coldrig lacks. A production history, customer base, public reviews, and a live trial all reduce buyer risk.
Where Coldrig's design is different
- One tenant credential. The agent does not need one API key per configured product.
- One managed state model. The intended customer experience hides the boundaries between registrar, mailbox, warmup, campaign, reply, and safety state.
- A smaller intent surface. The agent asks for outcomes such as
setup_infrastructureandlaunch_campaigninstead of operating several product object models. - No competing strategy agent. The user's chosen agent owns research and copy; Coldrig stays an execution layer.
Those are architectural advantages only when they match the owner's constraints. They do not compensate for an unavailable production feature.
Pricing: compare the all-in stack
Salesforge currently lists Pro at $48/month on monthly billing ($40/month annual) and Growth at $96/month ($80/month annual), with a 14-day trial. Mailforge, Infraforge, Primeforge, Leadsforge, and some other capacity are priced separately or by usage; calculate the exact set of products and mailbox type required. Coldrig's provisional formula is $49/month for the platform plus $10 per provisioned mailbox, with a five-mailbox minimum. That is $99 for 5 mailboxes, $149 for 10, $199 for 15, and $249 for 20. Paid real-sending activation is not live.
Bottom line
If the requirement is “run a real, agent-managed cold-email operation now,” the Forge Stack is the stronger and lower-risk recommendation. If the requirement is “test whether one stateful, agent-controlled abstraction can replace a multi-product integration,” Coldrig's free sandbox is worth a pilot. Re-evaluate the production verdict after Coldrig activates real sending, adds the event delivery the user needs, and has evidence from real operation.