Atlas entry
api.
14 capabilities over MCP. The inventory is incomplete. First atlas observation.
14 tools. 6 only read. 1 changes records. 7 have no reading.◇
AI reading, beta. side-effects v1 on claude-haiku-5.5, calibrated on MCP (golden-3). Only readings above a measured precision count. Counts are tools, not calls: no tool was called.
AI readings summary · beta
The service lists 14 tools, and 7 have AI readings.
voice_id_enroll is read as creating records.
6 tools are read as read-only, including voice_id_list_speakers, voice_id_identify and list_tts_voices.
transcribe_audio_pro: Returns a multilingual transcript of base64-encoded audio with word timestamps, confidence scores, and optional speaker labels.
Written by code from the AI readings on this page: it groups them and says nothing they do not.
What an agent can do here
AI readings, beta · Readings on 7 of 14 MCP tools.
Moves money
0
No tool read this way.
Deletes or cancels records
0
No tool read this way.
Sends messages to people
0
No tool read this way.
Changes permissions
0
No tool read this way.
Runs code
0
No tool read this way.
Should ask a person first
0
No tool read this way.
No counted reading 7 tools No reading was counted, so any effect above is possible for them: assess_pronunciation, check_stt_service, check_tts_service, synthesize_speech, transcribe_audio, transcribe_audio_pro, voice_id_verify.
What changed
First atlas observation. Atlas history
Look first
AI reading, beta: AI adds items here; it never removes one.
No tool is read as one of the six effects. A zero is a count, not a clearance.
Surfaces
5 protocols checked
| ARD | unreachable | No document retrieved |
|---|---|---|
| A2A | unreachable | 0 skills |
| MCP | Valid | 2025-11-25 · 14 tools · 0 resources retained · listing incomplete · 0 prompts retained · listing incomplete |
| OpenAPI | Not detected | 0 operations |
| WebMCP | Not detected | 0 declarative form tools · experimental |
Exact states and meaning
- ARDfailed
Exact state: unreachable · No document retrieved
The supported discovery location did not provide a usable response within the observation boundary.
An unreachable result does not prove the service is unavailable to every client or that it advertises no capabilities.
Not observed / not fetched
- A2Afailed
Exact state: unreachable · 0 skills
The supported discovery location did not provide a usable response within the observation boundary.
An unreachable result does not prove the service is unavailable to every client or that it advertises no capabilities.
Not observed / not fetched
- MCPobserved
Exact state: valid · Revision 2025-11-25 · 14 tools · 0 resources retained · listing incomplete · 0 prompts retained · listing incomplete
A usable protocol document was retrieved and validated.
Directly observed
- OpenAPIabsent
Exact state: not detected · 0 operations
No endpoint was established from the evidence and scope that InterfaceDelta inspected.
Not detected does not prove that no endpoint exists.
Not observed / not fetched
- WebMCPabsent
Exact state: not detected · 0 declarative form tools · experimental
No named declarative WebMCP form tools were observed in the origin root HTML.
Script-registered tools and other pages are not observed, so this does not mean the site has no WebMCP tools.
Not observed / not fetched
Capabilities
Showing 14 of 14.
| Tool and inputs | ◇ AI reading, beta · confidence | ||||||
|---|---|---|---|---|---|---|---|
| Read as none of the six 7 Not a clearance. | |||||||
| Checks whether the pronunciation scoring service is up, reporting its status, whether its model is loaded, and its API version.no inputs | Reads only0.80 | ||||||
| |||||||
| Checks whether the speech transcription service and its diarization pipeline are loaded and ready.no inputs | Reads only0.80 | ||||||
| |||||||
| Lists every English phoneme the pronunciation scorer can evaluate, with ARPAbet and IPA symbols, examples, and categories.no inputs | Reads only0.80 | ||||||
| |||||||
| Lists the text-to-speech voices on offer, with each voice's id, name, gender, accent and grade, plus the default voice ID.no inputs | Reads only0.88 | ||||||
| |||||||
| Registers a speaker's voice profile from a short audio clip, storing only an irreversible embedding, not the raw recording.audio*, group_id*, speaker_id* | Creates records0.85 | ||||||
| |||||||
| Returns enrolled speakers in a group ranked by voice similarity to a given audio clip, best match first.audio*, group_id*, top_k | Reads only0.80 | ||||||
| |||||||
| Lists voice profiles stored under a given group ID, with sample counts for each speaker.group_id* | Reads only0.87 | ||||||
| |||||||
| No counted reading 7 No reading was counted, so any effect is possible. | |||||||
| Assesses spoken English pronunciation from audio against reference text, returning overall, sentence, word and phoneme scores.audio_base64*, audio_format, text* | unknown | unknown | unknown | unknown | unknown | unknown | No reading |
| |||||||
| Checks whether the speech recognition service is up, reporting its status, whether the model is loaded, and the API version.no inputs | unknown | unknown | unknown | unknown | unknown | unknown | No reading |
| |||||||
| no inputs | unknown | unknown | unknown | unknown | unknown | unknown | No reading |
| |||||||
| Converts English text into a spoken WAV clip in one of twelve voices, returned as base64 audio with metadata.speed, text*, voice | unknown | unknown | unknown | unknown | unknown | unknown | No reading |
| |||||||
| Returns a text transcript of base64-encoded English speech audio, with optional per-word timestamps and confidence scores.audio_base64*, audio_format, include_timestamps | unknown | unknown | unknown | unknown | unknown | unknown | No reading |
| |||||||
| Returns a multilingual transcript of base64-encoded audio with word timestamps, confidence scores, and optional speaker labels.audio_base64*, diarize, language | unknown | unknown | unknown | unknown | unknown | unknown | No reading |
| |||||||
| Verifies whether an audio clip matches a previously enrolled speaker, returning a similarity score and match result.audio*, group_id*, speaker_id*, threshold | unknown | unknown | unknown | unknown | unknown | unknown | No reading |
| |||||||
Read as this effect No reading: unknown Not read as this effect◇ AI readings, beta. A zero is a count, not a clearance.
A2A skills 0+ observed
Inventory incomplete. The displayed count is a lower bound; missing items do not prove removal.
None recorded in this Passport.
MCP resources 0+ observed
Inventory incomplete. MCP resources could not be fully inventoried: ARD discovery did not resolve, so declarations made there were not read.
None recorded in this Passport.
MCP prompts 0+ observed
Inventory incomplete. MCP prompts could not be fully inventoried: ARD discovery did not resolve, so declarations made there were not read.
None recorded in this Passport.
OpenAPI operations 0+ observed
Inventory incomplete. OpenAPI operations could not be fully inventoried: ARD discovery did not resolve, so declarations made there were not read.
None recorded in this Passport.
The atlas shows each capability by its plain identifier, inputs and declared sign-in. Publisher descriptions, paths, links and source evidence are not published here.
The comparison with the previous observation is shown only to the account that recorded it.
Atlas history
1 observation
- 14 capabilities · MCP 2025-11-25 · MCP endpoint truncated · 2 incomplete inventory groupsFirst observation
Fingerprint and versions
- Fingerprint
c4e01e7b820a8b64c28e6648cb212e0b358b2827a75f7ddde9b7127e79c3ecc8(incomplete)- Passport schema
- 0.5.0, passport-semantic-v4
- Consistency
- MCP_PROMPTS_ADVERTISEMENT_LIST: indeterminate · MCP_RESOURCES_ADVERTISEMENT_LIST: indeterminate · MCP_TOOLS_ADVERTISEMENT_LIST: match · OPENAPI_MCP_EXPLICIT_MAPPING: not applicable
Each class compares an observation with the previous atlas observation. Runs of unchanged observations are grouped. Struck values are what the previous observation recorded; inserted values are what this one recorded.
Declared and observed
14 tools declare hints; the AI readings agree with all 7 that have a reading.
A disagreement is not a finding about the service: either side can be wrong. How we compare declarations and readings.
About this entry
com.brainiall/pronunciationListedOfficial MCP Registry · first listed 18 Sep 2026 · last listed 2 Nov 2026
Passive observations of a public interface listed in the Official MCP Registry. An observation does not verify security, ownership or compliance, and InterfaceDelta assigns no scores or rankings. Methodology · Report an inaccuracy · Ask for removal
Identity and snapshot
- Target key
sha256:1ce8d6d23432f5359a545700b3ddceaddff4cce5460a28cc30bf8206a2317e69- Locator SHA-256
615cace831ce529a77f365249999513e949a4ffb284b117e3f900046f1ae72f1- Snapshot generated
- 14 Nov 2026, 08:50 UTC
The entry is one exact observed target, not a real-world service identity, and implies no ownership. Endpoint paths are not published; the locator hash lets an operator confirm their own URL.