Discovery plumbing (secondary). This page makes the service findable by someone already searching; it does not replace pipeline work (page 00). Do it once; no paid tier, listing or upgrade (hard rule 6).
Prerequisite: Steps 1–2 (package + official registry entry live).
Three small submissions, one sitting. The single source of listing copy is
integrations/mcp/listing/descriptions.md (embedded below) — paste from it,
never freestyle. Note: PulseMCP is deliberately absent — their submissions
are paused (recheck ~2026-11-20).
apertureapi and
"claim" the server with the company GitHub account. Artifact: the Glama
server page URL showing it claimed.| Date | Venue | Submitted URL | Live URL |
|---|---|---|---|
| Glama | |||
| mcp.so | |||
| awesome-mcp |
Embedded verbatim from
docs/discovery/submission_drafts/glama_mcpso_claims.md— the in-repo file is the source of truth.
DRAFT ONLY. No account created, no claim started, no issue opened. The
operator submits. R9 Step 5, batched with the awesome-mcp-servers PR because
all three reuse one piece of copy.
Listing copy lives in integrations/mcp/listing/descriptions.md — the
one-liner, the two-sentence version, the 200-word body, the category and the
tag list, in priority order. Take them verbatim. Do not re-improvise per venue;
three venues drifting apart is how a claim stops matching the thing it claims.
Connection details live in integrations/mcp/listing/server.json — the
canonical name com.apertureapi/apertureapi, the package identifier, the
transport, and the environment variables. Quote from there rather than retyping.
What it is. The largest MCP index by a wide margin — 90,781 servers as of 2026-09-22. It auto-indexes open-source MCP servers from GitHub, so our listing may already exist without us doing anything. Only about 5,396 of those 90,781 listings are claimed, which is the whole opportunity: a claimed listing with real documentation is differentiated for almost no work.
Procedure
glama.ai/mcp/servers for ApertureAPI and for the repository URL. If
an auto-indexed entry exists, use the ownership claim flow on it. Do not
create a duplicate — a second entry competing with the auto-indexed one is
worse than either alone.descriptions.md: one-liner in the short field,
200-word body in the description, category Science & Research, tags in the
documented priority order trimmed from the end if the form caps them.Unverified, expect surprises: claim latency, whether claiming costs
anything, and Glama's moderation policy were all unverified in R9. If the claim
flow asks for domain or repository proof, the repository is public and the
apertureapi.com DNS zone is controllable — but confirm who holds each before
starting, not halfway through a verification challenge.
Verifiable artifact when executed: the Glama listing URL showing a claimed state with our description rendered.
What it is. A community aggregator, roughly 20,000+ servers (count is from trade press, not the site — treat as order-of-magnitude). Lowest-effort submission of any registry in the plan: one GitHub issue.
Procedure. The site states the mechanism verbatim: submit "by creating a new issue in our GitHub repository. Click the 'Submit' button in the navigation bar or visit our GitHub issues page directly. Please provide details about your server including its name, description, features, and connection information."
That sentence is the field list. Answer exactly those four.
Issue title
Submit: ApertureAPI — holographic reconstruction (no account required)
Issue body
**Name:** ApertureAPI
**Registry name:** com.apertureapi/apertureapi
**Repository:** https://github.com/startmotionmedia/ApertureAPI (server source in `integrations/mcp/`)
**Website:** https://apertureapi.com
**Description**
<paste the two-sentence version from integrations/mcp/listing/descriptions.md>
**Features**
- `reconstruct_hologram` — upload a hologram image (TIFF, PNG, NPY, JPG) and
get back reconstructed amplitude and phase, a height map, a 3-D surface as
`.obj`, rendered previews, and JSON provenance sidecars.
- `list_calibration_targets` — a catalog of inexpensive physical reference
objects with known geometry. Shoot one on your own rig, pass that image
alongside your hologram, and a digital-twin optimizer fits your true pixel
pitch and wavelength before reconstructing. This matters because a wrong
wavelength produces a confidently wrong reconstruction rather than an error.
- `get_service_health` — returns the commit, branch and environment digest of
the build that served the request, so a result can be pinned to exact code.
- Runs against a **no-account public endpoint by default**: no key to
provision before the first successful call.
- Every run returns a quality report stating its own pass/warn/fail band.
Metrics the system cannot establish — absolute height accuracy, per-pixel
uncertainty, repeatability — are returned as `not-yet-established` with a
null value rather than estimated.
**Connection information**
<paste the package identifier, transport and environment variables from
integrations/mcp/listing/server.json — do not retype them>
Verifiable artifact when executed: the mcp.so issue URL, and the live listing URL once it is triaged.
Ninety thousand listings is a phone book, not a distribution channel. Being
listed is table stakes. These registries matter because an agent can resolve
us there, not because a human will browse to us. The human path runs through the
public-apis row and the thing a stranger can click and run.
Embedded verbatim from
docs/discovery/submission_drafts/awesome_mcp_pr.md— the in-repo file is the source of truth.
punkpeye/awesome-mcp-servers pull requestDRAFT ONLY. No fork, no PR. The operator submits. R9 Step 5.
Sequencing: this comes after the official MCP Registry publish (R9 Step 1). The list links out to Glama, and Glama reads from the ecosystem — an entry that points at a published, resolvable server is a different proposition from one that points at a repository folder.
Listing copy is not in this file. It lives in
integrations/mcp/listing/descriptions.md, which is owned by the MCP server
work. Take the one-liner and the tags from there verbatim. If the copy needs to
change, change it there and re-derive — three venues reusing one source is the
point.
Read CONTRIBUTING.md in the repository first. R9 flagged its criteria as
unverified — they were not fetched during that research, and an awesome-list
maintainer rejecting a PR on a formatting rule nobody read is the most avoidable
failure mode here. Specifically check: the category taxonomy in use, whether an
emoji legend is required on each entry (this list has historically used marks
for language and for local/remote operation), sort order within a category, and
whether one PR per server is expected.
The list is organised by category. The plausible homes, in order:
Pick one. Do not add the entry twice; cross-posting across categories is how an awesome-list PR gets closed.
- [ApertureAPI](https://github.com/startmotionmedia/ApertureAPI) 🐍 ☁️ - Holographic reconstruction: amplitude, phase and 3-D surface from a single hologram image.
Mechanical notes:
integrations/mcp/listing/descriptions.md. Do not rewrite it here.README.md/CONTRIBUTING.md actually specifies, and drop them
entirely if the list does not use marks.Add ApertureAPI
Adjust to match the title convention in recent merged PRs on the repository.
Adds ApertureAPI to <category>.
An MCP server that gives an agent digital holographic reconstruction: hand it a
hologram image and it returns reconstructed amplitude, phase, a height map, a
3-D surface and JSON provenance sidecars.
Relevant to this list specifically:
- It runs against a **no-account public endpoint by default**, so there is no
key to provision before an agent's first successful call.
- It exposes a calibration-target catalog, because a wrong wavelength or pixel
pitch produces a confidently wrong reconstruction rather than an error — the
failure mode an agent cannot detect on its own is the one worth tooling for.
- Every run returns a quality report that states its own pass/warn/fail band,
and reports metrics it cannot establish as `not-yet-established` with a null
value rather than estimating them.
Published to the official MCP registry as `com.apertureapi/apertureapi`.
Source: `integrations/mcp/` in the linked repository.
CONTRIBUTING.md read and its actual rules applied over the guesses above.integrations/mcp/listing/descriptions.md exactly.name field in integrations/mcp/listing/server.json.Verifiable artifact when executed: the merged PR URL, plus the ApertureAPI entry visible in the rendered list under its category.
Embedded verbatim from
integrations/mcp/listing/descriptions.md— the in-repo file is the source of truth.
Written once here, reused across Smithery, Glama, mcp.so, PulseMCP and the official MCP registry. Do not re-improvise per venue — if a claim needs to change, change it here and re-derive.
Claims discipline applies: no number appears that is not backed by a committed artifact in this repository. Banned from all listing copy — NCC figures quoted as generic accuracy, "99.9 % error detection", any streaming fps number.
description field)Holographic reconstruction: amplitude, phase and 3-D surface from a single hologram image.
ApertureAPI turns a single hologram image into reconstructed amplitude, phase, and a 3-D surface, with provenance and a self-scoring quality report attached to every run. It works with no account — point an agent at an image and it gets a result archive back.
ApertureAPI MCP gives an agent digital holography as three tools.
Hand
reconstruct_holograman interference image — TIFF, PNG, NPY or JPG — and it returns a result archive: reconstructed amplitude and phase, a height map, a 3-D surface as.obj, rendered previews, and JSON sidecars carrying full provenance. It runs against the public no-account endpoint by default, so there is no key to provision before the first call.Optics are the hard part, and the server treats them that way. A wrong wavelength or pixel pitch yields a confidently wrong reconstruction rather than an error, so
list_calibration_targetsexposes a catalog of cheap physical reference objects with known geometry. Shoot one on your own rig, pass that image alongside your hologram, and a digital-twin optimizer fits your true pixel pitch and wavelength before reconstructing.Every run ships a quality report that states its own band — pass, warn or fail — and emits absolute height accuracy, per-pixel uncertainty and repeatability as
not-yet-establishedrather than inventing them.get_service_healthreturns the commit, branch and environment digest of the build that served you, so an agent can record exactly what code produced a result.
Primary category: Science & Research Secondary: Developer Tools / APIs (venues that allow a second category)
Tags — use in this order, trim from the end where a venue caps the count:
holography, microscopy, image-processing, 3d-reconstruction,
metrology, scientific-computing, computer-vision, optics
Rationale: holography and microscopy are the terms a domain user searches;
image-processing and 3d-reconstruction are the terms a generalist
developer searches. Both audiences have to find it, and the first four tags
cover both.
warn. If a venue's copy needs a number,
cite that artifact by name with its band attached, or cite nothing.traceability_status.si_traceable is false.Buyer pages to pair with this page: workflow fit. Pipeline stages and KPIs: page 00.