Agent-ready build vs MCP generators (2026)
If your API already has a clean OpenAPI spec and one tool per endpoint is close to what an assistant needs, a generator such as liblab or Fern may be all you need. Soxoa's agent-ready build is for products where the tools need designing around the jobs users actually do, with agent sign-in, confirmation for risky actions and directory listings, built in ten working days and handed over.
Soxoa publishes this page and sells the agent-ready build, one of the options compared here. Each vendor fact below was read on that vendor's own site on and links to its source. Facts we could not read on a vendor's site are left out. Product names belong to their owners.
At a glance
| Option | What it is | Published price |
|---|---|---|
| Soxoa agent-ready build | A production MCP server with tools designed for your product, OAuth agent sign-in with per-tool scopes, and registry and directory listings, built by John Arndt in your repository. | A fixed range, shown on the agent-ready page |
| liblab | MCP server generated and hosted from an OpenAPI spec | 100 MCP calls a month free (first year only), then $5 for every 100 calls. source |
| Fern | MCP server generated from an OpenAPI spec, hosted or self-run | Not stated on the pages we read |
| Mintlify | Search MCP server for a documentation site | Not stated on the pages we read |
| Your own engineers | Build an MCP server in house against the open protocol. | Engineering time |
The other options, from their own sites
liblab
- liblab says it can generate and host an MCP server directly from an OpenAPI spec, without the customer deploying or setting up infrastructure. source
- Its hosted MCP guide lists four authentication choices for how AI clients authenticate with the MCP server: API key, basic auth, bearer token, or none. source
- The guide describes the hosted service as an alternative to a local CLI option, which has its own guide. source
- Price: 100 MCP calls a month free (first year only), then $5 for every 100 calls. source
Pick liblab when: Your API already has a clean OpenAPI spec, one tool per endpoint is close enough to what an assistant needs, and you would rather pay by usage than for a build.
Checked on liblab.com.
Fern
- Fern's MCP generator builds a server from an OpenAPI spec and generates one tool per supported API operation. source
- By default the server is hosted by Fern at an address under fernmcp.dev; setting the delivery to zip instead downloads files the customer runs themselves with Node.js. source
- The Fern CLI and a Fern account are listed as prerequisites, and deploying a hosted server requires signing in with the CLI. source
Pick Fern when: You already use Fern for SDKs or docs, your engineers are comfortable owning a generated server, and a tool for every operation is what you want to expose.
Checked on buildwithfern.com.
Mintlify
- Mintlify generates a search MCP server for a documentation site and hosts it at the /mcp path of the site. source
- The tools it lists are search, reading the documentation files, and submitting feedback. source
- If the documentation site requires sign-in, users must sign in with OAuth before connecting, and the server returns only the content their user groups can access. source
Pick Mintlify when: Your docs already run on Mintlify and the goal is for assistants to answer questions from your documentation, not to take actions in your product.
Checked on mintlify.com.
When to pick another option
- Your API is already described by a clean OpenAPI spec, and a tool for every operation is what you want to expose: a generator is built for that job.
- The goal is for assistants to answer questions from your documentation, not to act in your product: a docs MCP server fits that job.
- Your team has engineers with time to design the tools, add OAuth and own the server: building in house keeps everything with them.
When the agent-ready build fits
- The assistant needs to do whole jobs in your product, so the tools have to be named, limited and described around those jobs rather than mirroring every endpoint.
- Agents must act as a signed-in user with a scope for each tool, and actions that spend money, delete data or send a message need confirmation in your product first.
- You want the server tested in ChatGPT, Claude and a coding agent, listed in the Official MCP Registry and directories, and handed over with the code in your repository and a runbook.
Details, the ten-day schedule and the price range are on the agent-ready page. The live MCP servers Soxoa runs on its own products are listed there too.
More comparisons Soxoa publishes are on the compare page, and the facts about Soxoa itself are on one page.
Vendor facts checked