# Vendor facets — Glama. The largest MCP directory (91,894 servers, 23,982 remote connectors) plus
# hosting and a gateway. A hosted deployment runs the provider's own code but is reached at
# glama.ai/endpoints/<profile>/mcp, Glama's domain behind a shared gateway, so it grades `platform`
# (0.25). The directory, inspector scores and glama.json claim are discoverability for Glama.
vendor: glama-ai
name: Glama
website: https://glama.ai
areas:
- mcp-hosting
registry_keys:
- glama
rubric_schema_version: 0.22.0
generated: '2026-09-25'
features_refreshed: '2026-09-25'
basis: capability
summary: >-
  Glama hosts MCP servers built from the provider's own GitHub repo, npm/PyPI package or Dockerfile, but
  every deployment is reached through the Glama Gateway at a glama.ai URL, so a provider that declares
  it earns `platform` credit (0.25 of 12 agent-readiness points) on mcp_server and nothing else. The directory
  listing, quality score and glama.json claim make a server findable on Glama; no Kin Score check reads
  them. Glama's own llms.txt and RFC 9727 api-catalog are Glama's surface, not something it serves for
  a provider.
features:
- id: hosting
  name: MCP server hosting on dedicated machines
  description: >-
    Deploys a server from the Glama registry, a GitHub repo (via glama.json or Dockerfile), or an npm/PyPI
    package or image, with health checks and encrypted env vars.
  source: https://glama.ai/mcp/hosting
  tier: paid
- id: gateway
  name: Glama MCP Gateway endpoints
  description: >-
    Every hosted deployment and any routed connector is served at https://glama.ai/endpoints/<connection-profile>/mcp
    with call logging, per-tool access control and managed OAuth 2.1.
  source: https://glama.ai/mcp/gateway
  tier: paid
- id: directory
  name: MCP directory of servers and connectors
  description: >-
    Indexes servers from GitHub, the official MCP Registry and submissions; scans, scores and ranks each
    one.
  source: https://glama.ai/llms.txt
  tier: all
- id: glama-json-claim
  name: glama.json claim file
  description: >-
    A repository commits glama.json naming GitHub maintainers to claim and maintain its directory listing.
  source: https://glama.ai/mcp/schemas/server.json
  tier: all
- id: connectors
  name: Remote MCP connector listings
  description: >-
    Lists remote MCP servers by their own native URLs (e.g. mcp.<service>.com/mcp) with OAuth/health attributes.
  source: https://glama.ai/mcp/connectors
  tier: all
- id: inspector
  name: Browser MCP Inspector
  description: Tests any MCP server in the browser including OAuth 2.1, DCR and PKCE; a client-side test
    tool.
  source: https://glama.ai/mcp/inspector
  tier: all
maps:
- feature: gateway
  check: mcp_server
  layer: agent_readiness
  grade: platform
  credit: 0.25
  provider_must: >-
    Declare the hosted server in its own mcp/ manifest. The endpoint the rubric probes is on glama.ai
    behind one shared gateway, so it reads as platform even though the tool code is the provider's own.
  note: >-
    Arguably closer to `templated` than a generated server because the provider wrote the tools; 0.22.0
    still grades a vendor-domain deployment `platform`, and the domain is what the rubric reads. Running
    the same code on the provider's own host is what earns more.
  points: 12
  baseline_pass_rate: 0.22
earns_nothing:
- feature: gateway
  check: protected_resource_metadata
  why: >-
    Managed OAuth lives in the Glama Gateway on glama.ai; nothing is served at /.well-known/oauth-protected-resource
    on the provider's host.
- feature: gateway
  check: dynamic_client_registration
  why: >-
    No registration_endpoint is published in the provider's own discovery document; Glama brokers credentials
    itself.
- feature: directory
  check: well_known_catalog
  why: >-
    Glama's api-catalog linkset names Glama's APIs; a directory listing is not a WellKnown/APICatalog
    pointer the provider declares.
- feature: glama-json-claim
  check: agent_skill_present
  why: >-
    glama.json is a maintainers list for claiming a listing; it is not an Agent Skill or any artifact
    a check reads.
- feature: connectors
  check: mcp_server
  why: >-
    A connector listing points at the provider's own remote server; whatever that server earns, it earns
    without Glama.
- feature: gateway
  check: mcp_endpoint_discoverable
  since: 0.23.0
  why: >-
    The hosted endpoint is glama.ai/endpoints/<connection-profile>/mcp, gated by per-profile tokens and
    private by default; nothing public fills in the profile, which 0.23.0 grades `templated` (0.0).
- feature: inspector
  check: auth_clarity
  why: >-
    The inspector tests a server's OAuth from the client side; it publishes nothing on the provider's
    surface.
out_of_reach:
  checks:
  - contract_present
  - spec_presence
  - llms_txt_published
  - agent_skills
  - agent_card
  - well_known_published
  - delegated_identity
  note: >-
    Glama runs and lists MCP servers; it creates no contract, llms.txt, Agent Skill, A2A card or provider-side
    .well-known document.
unscored_practice:
- feature: directory
  why: >-
    Glama's per-server security/quality scoring is a real third-party grade of MCP servers that no Kin
    Score check reads.
surface:
  agent_readiness:
    reachable: 3.0
    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://glama.ai/llms.txt
- https://glama.ai/mcp/connectors
- https://glama.ai/mcp/gateway
- https://glama.ai/mcp/hosting
- https://glama.ai/mcp/inspector
- https://glama.ai/mcp/schemas/server.json
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: 6825
  conditional_rows: excluded (they depend on what the API already does)
  composite_lift:
    median: 0.0
    p75: 0.0
    p90: 0.0
    max: 0.0
    mean_among_movers: 0.0
  agent_readiness_lift:
    median: 2.1
    p75: 2.2
    p90: 2.2
    max: 2.6
    mean_among_movers: 2.2
  facet_lift_median_among_movers: {}
  composite_band_moves: {}
  agent_readiness_band_moves:
    agent-aware -> agent-ready: 639
    agent-ready -> agent-native: 68
  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