BYOK for DiscoGen: Your Models, Your Search

BYOK for DiscoGen means you bring your own AI models and your own search providers. No more paying platform markup on the two most expensive parts of an AI research agent. You control the cost, the quality, and the provider.

Claygent just got expensive. If you have been watching the pricing changes, you already know. But here is what most people miss: the reason Clay could raise prices is that the alternatives are even more expensive. Every agent in this space depends on the same costly LLM built-in search. That dependency is the bottleneck, and it is the reason nobody undercut them.

Until now.

Why LLM Default Search Is the Problem

Every AI agent that uses a model’s built-in search capability is paying a hidden tax. The model calls its own search API, which is rate-limited, expensive, and opaque. You cannot control how many searches it runs. You cannot swap in a cheaper provider. You cannot even see what queries it executed.

This is the lever that lets platforms charge what they charge. The search cost is baked into the agent cost, and you have no say in it.

Claygent was the heart of Clay’s product because it combined search with company validation. But the search layer was never yours to control. That changes now.

Bring Your Own Search Providers

Serper and similar search APIs are delivering dramatically better price performance than any model’s default search. We are talking about an order of magnitude difference in cost for comparable or better results.

DiscoGen now lets you plug in your own search provider. Swap out the expensive default for Serper or any provider that fits your stack. You choose the search depth, the cost per query, and the provider.

This is not a minor optimization. This is the difference between research costing $30 per thousand records and costing $3.

Free Homepage Text for ICP Validation

The step that made Claygent indispensable was validating ICP fit against a company’s actual homepage content. That is the moment where you go from a list of names to a list of qualified targets.

Homepage text is now free with DiscoGen. No additional cost for Discover results records. Every company profile includes full homepage text at no charge, so your prompts are anchored with real company context instead of guessing from a name and a domain.

This is what separates a research agent from a glorified search wrapper. Context makes the prompt work. Without it, your agent hallucinates company details or returns generic nonsense.

API and MCP for Full Control

This is 2026. An AI research agent that only works inside one platform is a product limitation, not a feature.

DiscoGen is available through both API and MCP. Run it directly from Claude Code, integrate it into n8n workflows, or call it from any tool that speaks HTTP.

The screenshot at the top shows the real workflow: one prompt to the DiscoLike MCP finds US-based project management SaaS companies, another runs DiscoGen to research their pricing models and free trial availability. The results come back structured and ready to act on.

Create a target accounts list with one prompt. Validate with another. Add DiscoGen columns with a few more. No platform switching, no manual export-import loops.

What You Actually Pay

Processing prompts for 1,000 records:

$3–6 with top-tier models (Claude, GPT-5.3)
$1–2 with the best older models

No hidden fees. No platform markup on your own API keys. BYOK for both model and search providers, available on every plan including Starter.

Two more things that matter at scale:

Delegate costly bulk actions outside of Claude to preserve tokens and avoid flushing your context window. DiscoGen handles the heavy lifting so your agent session stays clean.
Bring a research agent to environments that don’t have one. If you are building in n8n or any API-based workflow, DiscoGen plugs in directly.

This Is Real

This post went live on April 1st. It reads like a joke. An AI research agent with BYOK models, BYOK search, free homepage text, MCP and API access, on every plan, for a few dollars per thousand records.

It is not a joke. It is live today.

Welcome to the moon orbit of GTM.

Frequently Asked Questions

What does BYOK mean for DiscoGen?

BYOK stands for Bring Your Own Key. With DiscoGen, you provide your own API keys for both the AI model (Claude, GPT, etc.) and the search provider (Serper, etc.). DiscoLike does not mark up these costs — you pay the providers directly at their published rates.

How does DiscoGen compare to Claygent on cost?

Claygent bundles model and search costs into platform pricing that you cannot break apart. DiscoGen lets you control both independently. The result is typically $3–6 per thousand records with top models versus significantly higher costs on Clay, especially after their recent pricing changes.

Can I use DiscoGen without Claude Code?

Yes. DiscoGen is available through both MCP and REST API. You can integrate it into any workflow tool (n8n, Make, custom scripts) or call it from any environment that supports HTTP requests. Claude Code via MCP is one option, not a requirement.

Is homepage text really free?

Yes. Homepage text is included at no additional cost for all Discover results records on every plan. This is the full homepage content that anchors your research prompts with real company context for ICP validation.

Related posts: