cloudparts

Speakeasy

Generates typed SDKs, Terraform providers, CLIs and MCP servers from an OpenAPI document, and now also sells a hosted MCP platform and an enterprise MCP gateway that governs which agents may call which tools.

What it is
Agent tools and integrations
Used for
Web APIs
Type
Commercial product
Made by
Speakeasy Development, Inc.
License
Proprietary service; CLI under the Elastic License 2.0; Gram under AGPL-3.0
Facts checked
13 October 2026
Website Source code

What it is

Speakeasy started as an SDK generator, and that is still how most developers meet it. You point its CLI at an OpenAPI 3.0 or 3.1 document; speakeasy quickstart asks whether you want an SDK or an MCP server and in which language, then speakeasy run validates the spec, generates the code and compiles it. Generators exist for TypeScript, Python, Go, Java, C#, PHP and Ruby (Rust and C++ are marked “coming soon”), each written to be idiomatic: the TypeScript SDK uses fetch, validates every model with Zod and offers optional retries and pagination, and the Python one is built on Pydantic models. Changes to names or examples go in OpenAPI Overlays so the original spec stays untouched, and the output is meant to be committed to GitHub and regenerated in CI. Free accounts can generate one SDK of up to 50 API methods.

The same spec can become an MCP server. speakeasy quickstart --mcp writes a TypeScript server, optionally with a Cloudflare Worker config, or Speakeasy hosts it: every operation becomes a tool, regenerated on each push, behind OAuth with your identity provider or Speakeasy as the authorization server, and every call logged with the customer and client. That hosted platform grew out of Gram, which Speakeasy open-sourced in 2025.

Since 2026 the company sells itself mainly as an “AI control plane” for enterprises: an MCP gateway with an approved registry of servers, access provisioned from Okta or Entra ID by role, inspection of prompts and tool calls for PII and prompt injection, and an audit trail of every call. Its public pricing page lists only an Enterprise plan for this, priced by tool-call requests.

Why teams pick it

  • One OpenAPI document drives SDKs in seven languages, Terraform providers, a CLI and an MCP server, so they stay in step with the API
  • Idiomatic output: typed models (Zod in TypeScript, Pydantic in Python), with optional retries and pagination, rather than a generic template
  • Overlays keep customizations outside the spec, so regeneration in CI doesn’t lose them
  • A path from a generated MCP server to a hosted one with OAuth, per-customer logs and OpenTelemetry export

Watch out for

  • Not open source: the CLI is under the Elastic License 2.0 and generation needs a Speakeasy account; the free tier covers one SDK of up to 50 methods
  • The output is only as good as the OpenAPI document, which often needs cleanup and overlays first
  • The company’s focus has moved to enterprise MCP governance, and that product has no public price
  • Its own pricing FAQ warns that model accuracy drops once a server exposes 30 to 40 tools, so a large API needs curation before it makes a good MCP server

Alternatives

What all agent tools and integrations have in common →Compare side by side →
ComposioCommercialA hosted tool layer for AI agents: one session per end user gives the model a few meta tools to search, authorize and run actions across more than 1,500 app toolkits, with Composio holding and refreshing the OAuth tokens.GlamaCommercialThe largest public directory of MCP servers, which builds and runs each open-source server in a Firecracker sandbox to record its tools, score their descriptions and flag risky behavior, plus hosting and a gateway for the servers you use.Model Context Protocol (MCP)Open sourceAn open protocol, built on JSON-RPC, that lets an AI application discover and call tools, read resources and use prompt templates from any server that speaks it, so one integration works across Claude, ChatGPT, Cursor, VS Code and other clients.OneRenamedCommercialA hosted connection layer for AI agents, formerly Pica: you connect each app once in your own account, and every agent reaches it through one MCP server or CLI, with per-action scopes enforced and logged on each call.OpenAPI SearchOpen sourceA small, open directory of OpenAPI descriptions for popular APIs, with an MCP server that turns any of them, or any spec URL, into a short plain-language summary an AI model can read before writing calls.

How it fits with other technologies

Generate an SDK or an MCP server from an OpenAPI document · bash

brew install speakeasy-api/tap/speakeasy

speakeasy quickstart        # sign in, give it an OpenAPI 3.x file or URL, pick SDK and a language
speakeasy quickstart --mcp  # or generate a TypeScript MCP server (optionally for Cloudflare Workers)

speakeasy run               # after the spec changes: validate, regenerate and compile

See it in a real stack

No architecture uses Speakeasy yet. Drop it into the playground and wire it up yourself.Open playground →

Worth exploring

WebsiteOpenAPI.ToolsPhil Sturgeon and APIs You Won’t Hate, with community contributorsAn open-source directory of more than 300 tools for working with OpenAPI descriptions, sorted into 23 categories such as mock servers (about 25, including Prism, Microcks, WireMock and Specmatic), SDK generators, linters, documentation, gateways, testing and monitoring, and filterable by language and by which OpenAPI versions each supports. Entries arrive by pull request on GitHub; sponsors get priority placement and a badge in their categories, so the order isn’t a ranking.
All resources →
Where it came from2022–2026

Sagar Batchu, Speakeasy’s co-founder and CEO, announced the company in June 2023 with $11 million in pre-seed and seed funding from backers including GV and Quiet Capital. The pitch was the “final mile” of API integration: companies like Stripe, Plaid and Twilio win users with polished SDKs, but most API teams cannot afford to build and maintain them in many languages, so Speakeasy would generate managed SDKs and Terraform providers from the API’s own description. By launch it said it had generated 4,000 SDKs across 8 languages.

In October 2024 it raised a $15 million Series A led by FPV Ventures, describing its aim as ending “API embarrassment”: stale docs, accidental breaking changes and SDKs that lag the API. As agents arrived, the same generators began emitting MCP servers, and in September 2025 Speakeasy put Gram into public beta, an open-source platform for curating API endpoints into a small set of well-described tools and hosting the resulting MCP servers.

In 2026 Speakeasy repositioned around governing AI use inside companies, publishing an “AI control plane” reference in April and selling an MCP gateway for registries, role-based access, inspection and audit. SDK generation continues alongside, with CLI releases still shipping every few days.

Why it’s built this way

Generate from OpenAPI rather than ask teams to adopt a new schema
Most API teams already have an OpenAPI description, so making it the single source let Speakeasy produce SDKs, Terraform providers and later MCP servers without changing how the API is built.
A hand-written generator per language
Speakeasy built each generator with language experts, using Zod in TypeScript and Pydantic in Python, because SDKs that feel foreign to a language community go unused.
Curate tools instead of exposing every endpoint
Gram’s launch post argued that MCP servers fail when they expose too many tools without context, so it turns a 600-endpoint API into a focused set of 5 to 30 tools.

Timeline

  1. 2022

    speakeasy-api/speakeasy CLI repository created in September

  2. 2023

    Public launch with $11 million in pre-seed and seed funding (June)

  3. 2024

    $15 million Series A led by FPV Ventures (October)

  4. 2025

    Gram, an open-source MCP server platform, in public beta (September)

  5. 2026

    AI control plane reference and MCP gateway (April)

Sources11 · facts checked 13 October 2026

Improve this page on GitHub