# Vendor facets — Speakeasy. Generates SDKs, Terraform providers, CLIs and MCP servers from an
# OpenAPI, and hosts MCP servers through Gram. The headline: its lift is SDK and CLI pointers plus
# one rubric-visible contract artifact, the OpenAPI Overlay that carries its code samples. It
# publishes no docs site of its own, and its linter reports gaps without fixing them.
vendor: speakeasy
name: Speakeasy
website: https://www.speakeasy.com
areas:
- sdk-generation
registry_keys: []
rubric_schema_version: 0.22.0
generated: '2026-09-25'
features_refreshed: '2026-09-25'
basis: capability
summary: >-
  Speakeasy moves Developer Ergonomics through SDKs in seven languages and a generated Go CLI, and each
  counts only once the provider declares it in apis.yml. It is one of the few vendors whose output is
  an OpenAPI Overlay (the x-codeSamples code-samples overlay), which scores overlay_published if the provider
  publishes it. Those code samples are SDK snippets, not request/response examples, so they earn nothing
  on examples_present and only the link fallback on openapi_examples. Its MCP servers, Terraform providers
  and retry logic look like the customer's own work, so the customer surface carries almost no Speakeasy
  fingerprint.
features:
- id: sdk-generation
  name: SDK generation
  description: >-
    Idiomatic SDKs for TypeScript, Python, Go, Java, C#, PHP and Ruby (Rust and C++ listed as coming soon).
  source: https://www.speakeasy.com/docs
  tier: paid
- id: sdk-retries
  name: SDK retries driven by x-speakeasy-retries
  description: >-
    Backoff retries configured in the OpenAPI. When an endpoint already defines an idempotency header,
    retries reuse the same key.
  source: https://www.speakeasy.com/docs/sdks/customize/runtime/retries
  tier: all
- id: code-samples-overlay
  name: Code samples delivered as an OpenAPI Overlay
  description: >-
    Generates x-codeSamples usage snippets for every operation as an Overlay document (code-samples.yaml)
    that docs platforms such as Scalar, Mintlify, ReadMe and Redocly render.
  source: https://www.speakeasy.com/docs/sdks/sdk-docs/code-samples/generate-code-samples
  tier: all
- id: code-samples-api
  name: Code Samples API
  description: An authenticated API that serves generated snippets. Enterprise only.
  source: https://www.speakeasy.com/docs/sdks/sdk-docs/code-samples/code-samples-api
  tier: enterprise
- id: sdk-docs-embedding
  name: SDK snippets embedded in partner docs
  description: >-
    Speakeasy does not host an API reference. It adds snippets to the spec for a third-party docs platform
    to render.
  source: https://www.speakeasy.com/blog/release-sdk-docs
  tier: all
- id: cli-generation
  name: CLI generation
  description: >-
    A Go (Cobra) CLI that wraps the generated Go SDK, with GoReleaser configs and install scripts, and
    an agent mode that ships llms.txt and a skills.md for the CLI.
  source: https://www.speakeasy.com/blog/release-cli-generation
  tier: unknown
- id: mcp-generation
  name: Standalone MCP server generation
  description: >-
    A TypeScript MCP server generated from the OpenAPI, distributed as an npm package, an MCPB bundle
    or a Cloudflare Worker config.
  source: https://www.speakeasy.com/docs/standalone-mcp/build-server
  tier: all
- id: gram-hosting
  name: Gram hosted MCP servers
  description: >-
    Hosts MCP toolsets generated from an uploaded OpenAPI, with a custom slug and, on request to support,
    a custom domain.
  source: https://www.speakeasy.com/docs/gram/host-mcp/deploy-mcp-server
  tier: paid
- id: terraform-generation
  name: Terraform provider generation
  description: Generates a Terraform provider from x-speakeasy-entity annotated OpenAPI operations.
  source: https://www.speakeasy.com/docs/terraform/create-terraform
  tier: paid
- id: openapi-linter
  name: OpenAPI linter
  description: >-
    90+ rules in five rulesets (speakeasy-recommended, speakeasy-generation, speakeasy-openapi, openapi-standard,
    owasp), configured in .speakeasy/lint.yaml. It reports issues and does not fix them.
  source: https://www.speakeasy.com/docs/sdks/prep-openapi/linting
  tier: all
- id: speakeasy-skills
  name: Speakeasy Agent Skills
  description: >-
    21 skills that teach coding agents to use Speakeasy (specs, SDKs, Terraform). They are not generated
    for the customer's API.
  source: https://www.speakeasy.com/docs/speakeasy-reference/skills/
  tier: open-source
maps:
- feature: sdk-generation
  check: sdk_count_1
  layer: composite
  provider_must: Publish a generated SDK and declare it as an SDK entry in apis.yml.
  catalog_pass_rate: 0.148
  facet: developer_ergonomics
  points: 3
  baseline_pass_rate: 0.413
- feature: sdk-generation
  check: sdk_count_3
  layer: composite
  provider_must: Publish three or more language SDKs and declare each as its own SDK entry.
  catalog_pass_rate: 0.018
  facet: developer_ergonomics
  points: 4
  baseline_pass_rate: 0.043
- feature: cli-generation
  check: cli_present
  layer: composite
  provider_must: Release the generated CLI and declare it as a CLI entry.
  note: CLI generation is early-stage (alpha in the docs nav, beta in its own docs page).
  catalog_pass_rate: 0.063
  facet: developer_ergonomics
  points: 3
  baseline_pass_rate: 0.19
- feature: code-samples-overlay
  check: overlay_published
  layer: composite
  provider_must: >-
    Publish the code-samples Overlay document where the catalog harvests overlays/. By default it sits
    in the Speakeasy registry and the provider's workflow, not on the provider's public surface.
  catalog_pass_rate: 0.0
  facet: contract_governance
  points: 4
  baseline_pass_rate: 0.0
- feature: code-samples-overlay
  check: examples_published
  layer: composite
  provider_must: Declare the page or repository carrying the generated snippets as a CodeSamples entry.
  catalog_pass_rate: 0.052
  facet: contract_quality
  points: 4
  baseline_pass_rate: 0.127
- feature: code-samples-overlay
  check: openapi_examples
  layer: agent_readiness
  grade: documented
  partial: true
  partial_note: >-
    x-codeSamples are SDK usage snippets, not request/response examples. The dimension reads examples
    first, so Speakeasy reaches only the Code Samples link fallback (0.25).
  points: 7
  baseline_pass_rate: 0.36
- feature: cli-generation
  check: agent_skills
  layer: agent_readiness
  credit: 0.25
  provider_must: Generate a CLI and publish its skills file where the catalog can find it.
  note: >-
    The skills.md is generated from the CLI's commands, not written by the provider, so it is the derived
    class.
  points: 5
  baseline_pass_rate: 0.236
- feature: mcp-generation
  check: mcp_server
  layer: agent_readiness
  grade: documented
  provider_must: >-
    Publish and declare the server. An npm or MCPB package run locally cannot be probed. Reaching `verified`
    means deploying it on a host the provider operates (a Cloudflare Worker, or Gram on a custom domain).
  note: >-
    Tools are generated per operation from the provider's own OpenAPI, so their shape should not cluster
    with other customers' as `platform`. A Gram server left on Gram's default URL is not on the provider's
    surface.
  points: 12
  baseline_pass_rate: 0.22
- feature: openapi-linter
  check: operations_summary_coverage
  layer: composite
  conditional: true
  condition: The linter reports missing summaries. The provider has to write them and republish.
  catalog_pass_rate: 0.879
  facet: contract_quality
  points: 6
  baseline_pass_rate: 0.933
  saturated: true
  saturated_note: >-
    93% of providers with a contract, docs and a reference already earn this; the vendor cannot move it
    for most of its buyers.
- feature: openapi-linter
  check: operations_operation_ids
  layer: composite
  conditional: true
  condition: >-
    Only when the provider fixes what the generation ruleset flags and publishes that contract. The linter
    does not rewrite the document.
  catalog_pass_rate: 0.827
  facet: contract_quality
  points: 4
  baseline_pass_rate: 0.918
  saturated: true
  saturated_note: >-
    92% of providers with a contract, docs and a reference already earn this; the vendor cannot move it
    for most of its buyers.
- feature: openapi-linter
  check: error_responses_documented
  layer: composite
  conditional: true
  condition: Only if the API really returns documented 4xx bodies. Linting cannot invent an error schema.
  catalog_pass_rate: 0.367
  facet: contract_quality
  points: 4
  baseline_pass_rate: 0.621
- feature: openapi-linter
  check: credentials_not_in_query
  layer: composite
  conditional: true
  condition: >-
    The owasp ruleset flags query-string credentials. Clearing it means changing how the API authenticates,
    not the document.
  catalog_pass_rate: 0.845
  facet: contract_quality
  points: 3
  baseline_pass_rate: 0.891
earns_nothing:
- feature: code-samples-overlay
  check: examples_present
  why: x-codeSamples are SDK snippets. The check counts request or response examples on operations.
- feature: sdk-retries
  check: idempotency
  layer: agent_readiness
  why: >-
    Retries reuse an idempotency header only when the contract already declares one. The SDK never adds
    one.
- feature: speakeasy-skills
  check: agent_skill_present
  why: The skills teach an agent to use Speakeasy. They describe nothing on the customer's API.
- feature: openapi-linter
  check: rules_present
  why: >-
    A .speakeasy/lint.yaml is private workflow configuration. The check reads a ruleset the provider publishes.
out_of_reach:
  checks:
  - portal_present
  - documentation_present
  - api_reference_present
  - llms_txt_published
  - change_log_present
  - status_page_present
  - rate_limits_documented
  - well_known_published
  - auth_clarity
  note: >-
    Speakeasy hosts no docs site; it feeds snippets to Scalar, Mintlify, ReadMe or Redocly. The llms.txt
    it generates belongs to a generated CLI, not to the provider's docs root.
unscored_practice:
- feature: terraform-generation
  why: A Terraform provider is a real, deployed integration surface that no check reads.
- feature: gram-hosting
  why: >-
    Gram's AI control plane (policy and identity for agents and MCP) governs consumption, not the provider's
    published surface.
- feature: sdk-retries
  why: The x-speakeasy-retries extension records retry behaviour in the OpenAPI, and no check reads it.
surface:
  contract_quality:
    reachable: 21.0
    total: 211
  contract_governance:
    reachable: 4.0
    total: 33
  developer_ergonomics:
    reachable: 10.0
    total: 42
  agent_readiness:
    reachable: 7.8
    total: 139
hard_rule: >-
  A model, not a score. Adopting this vendor changes a provider's Kin Score only when the provider publishes
  the resulting artifacts on its own surface; nothing here writes a score, and no sponsorship or partnership
  can.
method: searched
source:
- https://www.speakeasy.com/blog/release-cli-generation
- https://www.speakeasy.com/blog/release-sdk-docs
- https://www.speakeasy.com/docs
- https://www.speakeasy.com/docs/gram/host-mcp/deploy-mcp-server
- https://www.speakeasy.com/docs/sdks/customize/runtime/retries
- https://www.speakeasy.com/docs/sdks/prep-openapi/linting
- https://www.speakeasy.com/docs/sdks/sdk-docs/code-samples/code-samples-api
- https://www.speakeasy.com/docs/sdks/sdk-docs/code-samples/generate-code-samples
- https://www.speakeasy.com/docs/speakeasy-reference/skills/
- https://www.speakeasy.com/docs/standalone-mcp/build-server
- https://www.speakeasy.com/docs/terraform/create-terraform
measured:
  cohort:
    method: vendors-catalog.json detections (CNAME / header / URL shape / markup), never a name match
    detected: 0
    in_baseline: 0
  control:
    basis: providers earning contract_present + documentation_present + api_reference_present, minus the
      cohort
    n: 5216
  metric: >-
    cohort_pct / control_pct = mean share of the check's points earned (derived and platform credit weighted),
    x100
  measured_on: '2026-09-25'
  status: 'not measurable: 0 detected customers clear the baseline (need 20)'
simulation:
  simulated_on: '2026-09-25'
  rubric: 0.23.0
  population: providers publishing a contract (contract_present earned), replayable exactly
  providers: 8977
  providers_unreplayable: 987
  providers_moved: 8976
  conditional_rows: excluded (they depend on what the API already does)
  composite_lift:
    median: 6.9
    p75: 6.9
    p90: 6.9
    max: 7.6
    mean_among_movers: 6.1
  agent_readiness_lift:
    median: 4.4
    p75: 5.6
    p90: 5.6
    max: 6.5
    mean_among_movers: 4.3
  facet_lift_median_among_movers:
    contract_quality: 2.7
    contract_governance: 12.1
    developer_ergonomics: 23.8
  composite_band_moves:
    thin -> developing: 2016
    developing -> strong: 1224
    strong -> exemplar: 400
    emerging -> thin: 361
    minimal -> emerging: 3
  agent_readiness_band_moves:
    agent-aware -> agent-ready: 1950
    agent-ready -> agent-native: 151
  method: >-
    each provider's own kin/checks file, the vendor's maps at their stated credit, the scorer's composite
    formula; from -> to, nothing written