What to prepare
- A valid official registry identity with a verified existing record or an eligible official-registry publication in the same launch.
- A reproducible source package or reachable remote endpoint that can complete Glama’s evaluation.
- The service operator’s authenticated Glama discovery integration and an available verification worker.
A profile, an evaluation, and a discoverable listing
These are distinct states. A source profile can remain on Glama when a reproducible build fails, while search distribution is withheld. Static repository inspection identifies preparation issues at a pinned commit; only Glama can establish its build, runtime introspection and evaluation results. A successful page fetch does not establish discoverability.
The worker checks the exact repository or endpoint identity, an evaluated tool-definition score, and the matching provider record in discovery results. A remote connector also needs a completed healthy test and must not be deprecated. Unknown or incomplete evaluation remains pending.
Publish through the documented ingestion route
Glama describes its catalog as a superset of the official MCP Registry. MCP Server List can publish an eligible registry record and follow the resulting Glama ingestion, or use a verified existing registry record as the source. Registry publication is evidence of the upstream action; it does not establish that Glama has finished its own checks.
A launch can therefore show awaiting_ingestion after the registry action succeeds. If evaluation or discovery cannot be confirmed within the supported checking window, the record remains unresolved rather than becoming a fabricated listing.
Ownership management remains separate
Glama’s direct repository submission and claim flows use the maintainer’s authenticated identity. Personal repositories can use GitHub authentication; organization repositories need confirmed maintainers in root glama.json for ownership claims. This file is ownership metadata, not a submission receipt or runtime evaluation. Remote connector claims use their own domain or eligible GitHub verification.
A root Dockerfile can make the intended build reproducible, but its presence does not prove that Glama built or introspected the server. Glama may infer a Dockerfile or use its Admin build configuration; absence alone is not a documented rejection. Confirm the actual build and captured tools, and keep sandbox credentials private. Never publish a session, service API key or production connector credential in an ownership file.
{
"$schema": "https://glama.ai/mcp/schemas/server.json",
"maintainers": ["your-github-username"]
}Submission workflow
- Check for an existing exact repository or connector record during the free preflight.
- Approve the official-registry publication and dependent Glama verification scope where a new registry record is needed.
- After verified payment, let the worker publish or confirm the registry record and wait for Glama ingestion.
- Confirm evaluated discovery evidence; a retained profile or guessed badge URL is not completion.
Blockers to resolve before review
Use these as preparation checks. The destination makes its own acceptance decision.
- Repository metadata cannot be fetched or the repository is not a supported source.
- The server does not start or cannot answer introspection requests in the indexing environment.
- Ownership metadata has invalid JSON or names the wrong GitHub maintainer.
Primary sources
Checked 2026-10-29. Follow the destination’s current instructions if its requirements change.