Onepin
Server Details
AI Voice Production Agent: picks from 30+ TTS models, fixes pronunciation, validates every line.
- Status
- Healthy
- Uptime
- 100.0% over the last 24 hours
- OAuth
- Works in Glama
- Last Tested
- Transport
- Streamable HTTP · MCP 2025-11-25
- URL
- Repository
- podonos/onepin-mcp
- GitHub Stars
- 0
- ChatGPT
- –
- Claude
- –
TDQS
Scored across 39 tools
The set is organized into clear families (workflows, runs, uploads, dictionary, voices), and descriptions actively route the agent ("prefer this over create_workflow", "use this over update_workflow"), which helps a lot. However, four tools create workflows (create_workflow, create_workflow_from_intent, create_audio_workflow_from_intent, create_workflow_from_template) and two edit them (patch_workflow, update_workflow), with configure_quality_checks also overlapping patch_workflow, so boundaries must be learned from prose rather than names.
Every tool is verb_noun in snake_case (get_run_status, list_templates, validate_workflow_definition, cancel_run), including modifiers like _from_intent and _from_template. No camelCase, no bare nouns, no vague verbs like process or handle.
Thirty-nine tools is heavy for one server, and some pairs are near-duplicates that a parameter would fold (upload_audio_file/upload_script_file, finish_audio_upload/finish_script_upload, list_audio_files/list_script_files). The domain is genuinely broad — workflow authoring, run lifecycle, dictionary, voices, workspaces, docs — so most tools earn a place, but the surface is at the over-scoped end.
Coverage is unusually thorough: workflow create/read/update/patch across multiple authoring paths, full run lifecycle (estimate, run, poll, outputs, download, cancel, list), dictionary CRUD plus scanning and IPA suggestion, upload two-phase flows, voices, providers, and workspaces. The main gap is workflow deletion/archival — workflows can only be created and modified, never removed, and there is no cleanup path for uploaded files.
Publisher details
- Operator
- Podonos
- Operator website
- https://onepin.ai
- Vendor relationship
- First-party
- Documentation
- https://docs.onepin.ai
- Privacy policy
- Not provided
- Trust center
- Unknown
- Restrictions
- A free Onepin account is sufficient to connect; no paid plan, admin approval, or allowlisting is required. Sign-in is via OAuth 2.0 with Dynamic Client Registration, so no custom OAuth app has to be created. No regional restrictions. Running a workflow consumes credits — a paid plan is required once the free monthly credits are used up.
Available Tools
39 toolscancel_runCancel runADestructiveIdempotentInspect
Stop a run that is still in flight. Permanent — a cancelled run cannot be resumed.
Credits already spent are not refunded and finished lines are not delivered. Without confirm=true nothing is cancelled: you get the run's current state to show the user. Already-finished runs are returned unchanged.
Inputs: run_id (required), workflow_id (required), confirm. Returns: run_id, status, cancelled, next_step, +4 more
configure_quality_checksConfigure quality checksADestructiveIdempotentInspect
Change the quality policy on voice generators a saved workflow already has.
The server finds each branch and rewires every retry edge; you do not draw them. Use target to scope to ids or locales, and patch_workflow for a threshold on one node. Call without confirm for an edit_token, then again with confirm=true.
Inputs: checks (required), workflow_id (required), target, confirm, +4 more. Returns: name, note, valid, errors, +15 more
create_audio_workflow_from_intentBuild an audio workflowAInspect
Build a workflow around audio the user already has. The server draws it.
Give the uploaded clips with their languages and at least one check; the clips feed the checks, then an Export. A check drops its failed clips unless you wire its fail to an Export on the canvas. Call again with confirm=true.
Inputs: name (required), checks (required), audio_inputs (required), confirm, +2 more. Returns: name, nodes, valid, errors, +10 more
create_workflowCreate workflowAInspect
Create a workflow from a definition validate_workflow_definition approved.
Needs that call's validation_token, for these exact definition bytes. Without confirm=true it creates nothing and returns what to show the user, or a workflow you already have that does this (force=true skips the check). Returns workflow_id.
Inputs: name (required), definition (required), validation_token (required), force, +1 more. Returns: name, created, next_step, console_url, +5 more
create_workflow_from_intentBuild a standard workflowAInspect
Build the ordinary workflow for the languages you name. The server draws it.
Prefer this over create_workflow: send the languages with their voices and ONE script route; the server decides the nodes, ports, translator and quality checks. Author a definition only for a shape this refuses. Call again with confirm=true.
Inputs: name (required), speaks (required), source_language (required), force, +8 more. Returns: name, nodes, valid, errors, +10 more
create_workflow_from_templateCreate workflow from templateAInspect
Clone a template into this workspace as a runnable workflow.
Call only after the user picked THIS template — asking for audio is not picking one. Without confirm=true nothing is cloned: you get the pipeline, voices and quality gates the template already decided, to show them. force=true skips the reuse check.
Inputs: template_id (required), name, force, confirm. Returns: name, cloned, reused, next_step, +6 more
delete_dictionary_entryDelete dictionary entryADestructiveInspect
Remove one entry from the workspace pronunciation dictionary. It cannot be restored, only saved again.
Inputs: entry_id (required). Returns: deleted, entry_id, next_step
download_runDownload run audioAIdempotentInspect
Get one ZIP link containing every audio file a completed run produced.
Use this instead of downloading lines one at a time. Only completed runs can be exported; an unfinished, failed or cancelled run is refused. Returns url, filename and expires_at — the link stops working at expires_at, so fetch it now.
Inputs: run_id (required), workflow_id (required). Returns: url, filename, next_step, expires_at
estimate_runEstimate run costARead-onlyInspect
Price a run before starting it. Creates nothing, charges nothing, asks nothing.
Returns credits (expected/min/max), current_balance, can_run, a breakdown, alternatives and console_url. Nobody was asked which workflow this is — confirm it and the price with the user before anything spends.
Inputs: workflow_id (required), script_text, source_language. Returns: priced, can_run, breakdown, next_step, +13 more
find_voicesFind voicesARead-onlyInspect
Search for voice candidates — the ONLY tool that may search or recommend voices.
Call before preview_voice unless the user named a valid catalog_voice_id. Returns at most three, vetted but not ranked, with no audio: preview_voice each to compare, then let them say which is theirs. Never pick a voice for the user.
Inputs: age, limit, model, style, +7 more. Returns: voices, next_page, next_step, matched_by, +2 more
finish_audio_uploadFinish an audio uploadAIdempotentInspect
Register an audio clip whose bytes you already PUT to an upload_url.
Returns its size — never the audio. Pass the upload_id to a source_audio node line as audio_upload_id, with that line's language and (for Word Accuracy/Pronunciation) script.
Inputs: upload_id (required), workflow_id. Returns: lines, filename, next_step, upload_id, +4 more
finish_script_uploadFinish a script uploadAIdempotentInspect
Register a file whose bytes you already PUT to an upload_url.
Returns its character and line counts and detected language — never its text. Pass the upload_id to a source_script node as input_type="file", upload_ids=[upload_id].
Inputs: upload_id (required), workflow_id. Returns: lines, filename, next_step, upload_id, +4 more
get_balanceGet credit balanceARead-onlyInspect
Credits remaining, used and reset date for this workspace.
Answers "how many credits are left" without pricing a run. Returns the spendable balance, this billing period's usage against quota, and when credits next reset. What a specific run will cost, and whether the balance covers it, comes from estimate_run.
Inputs: no named parameters. Returns: used, quota, percent, next_step, +4 more
get_node_infoGet node infoARead-onlyInspect
The node types you can use, or one node's config schema and ports.
No argument: every node type available to this key, with its category. With node_type: that node's config_schema, inputs and outputs — read it before writing its config. A generator's voice_map needs real voices: call find_voices first.
Inputs: node_type. Returns: type, nodes, inputs, outputs, +7 more
get_run_outputsGet run outputsARead-onlyInspect
List the audio a run produced, one entry per output node.
Call once get_run_status reports completed. Nothing is played here — ask the user before you play anything. Each line carries takes[] (one audio_url each), normalized, and delivery_passed. The URLs are short-lived — re-call this tool rather than caching them.
Inputs: run_id (required), workflow_id (required). Returns: run_id, credits, outputs, next_step, +2 more
get_run_statusGet run statusARead-onlyInspect
Check how far a run has got.
Poll this after run_workflow until status is completed, failed or cancelled; do not start a second run while one is in flight. Returns status, finished_steps, total_steps, has_export, error, started_at, completed_at.
Inputs: run_id (required), workflow_id (required). Returns: error, status, next_step, has_export, +5 more
get_workflowGet workflowARead-onlyInspect
Fetch one workflow as a markdown page describing what its graph does.
Call after list_workflows. markdown is written to show the user as-is: pipeline,
nodes, voices, gates. Pass markdown=false for structured fields only, or
include_definition=true for the editable graph update_workflow takes.
Inputs: workflow_id (required), markdown, include_definition. Returns: name, voices, summary, markdown, +6 more
list_audio_filesList audio filesARead-onlyInspect
The audio clips a workflow can already read, newest first.
Call it after the user uploads one in the browser, to learn its upload_id — never guess one. Pass an upload_id to a source_audio node line as audio_upload_id.
Inputs: workflow_id (required), limit. Returns: count, files, next_step, workflow_id
list_dictionary_entriesList dictionary entriesARead-onlyInspect
The workspace pronunciation dictionary: words with their IPA and category, one locale or a search.
With neither locale nor search you get the locales that have entries, with counts.
Inputs: limit, locale, offset, search. Returns: entries, locales, next_step
list_providersList providersARead-onlyInspect
The speech providers you can use, or one provider's models.
No argument: every provider with how many models it has. With provider: that provider's models, each with how many catalog voices it has. Voices themselves come from find_voices.
Inputs: provider. Returns: models, provider, next_step, providers
list_runsList runsARead-onlyInspect
The recent runs of one workflow, newest first — how a lost run_id is found again.
Use this when the run_id run_workflow returned is no longer to hand, or to check whether one is still in flight before starting another. Returns run_id, status, step progress, credits and timestamps.
Inputs: workflow_id (required), limit. Returns: runs, next_step, active_run_id
list_script_filesList script filesARead-onlyInspect
The script files a workflow can already read, newest first.
Call it after the user uploads one in the browser, to learn its upload_id — never guess one. Each row carries the filename, its character and line counts, and a download link.
Inputs: workflow_id (required), limit. Returns: count, files, next_step, workflow_id
list_templatesList templatesBRead-onlyInspect
Browse the template gallery — ready-made workflows, and each is only a suggestion.
Tell the user which one would fit and ask whether to clone it or build a new workflow from operators and validations; asking for audio is not choosing a template. Returns template_id, name, description, category, languages, voices.
Inputs: limit, search, category. Returns: next_step, templates
list_workflowsList workflowsARead-onlyInspect
List this workspace's workflows, most recently updated first.
Check here before building anything new — one of these may already do the job. Every workflow_id another tool needs comes from here; never guess one. Returns workflow_id, name, description, languages, voices, and run stats.
Inputs: limit, search. Returns: next_step, workflows, workspace_slug
list_workspacesList workspacesARead-onlyInspect
List the workspaces you can work in, and which one is currently selected.
Inputs: no named parameters. Returns: count, workspaces, selected_workspace_id
patch_workflowPatch workflowADestructiveIdempotentInspect
Change part of a saved workflow. Everything the operations skip is kept.
Use this over update_workflow for an ordinary edit: a threshold, a config key, a node added or removed, a connection redrawn. Call without confirm for an edit_token, then again with the SAME patch and confirm=true. Voices: set_workflow_voice.
Inputs: patch (required), workflow_id (required), confirm, edit_token. Returns: name, note, valid, errors, +14 more
preview_pronunciationPreview pronunciationARead-onlyInspect
Speak one IPA string so the user can hear it before it is saved. Returns a short-lived WAV link.
One string per call, and only one the user asked to hear.
Inputs: ipa (required), locale (required). Returns: ipa, cached, audio_url, next_step
preview_voicePreview voiceARead-onlyInspect
Play ONE voice from find_voices' candidates — call once per candidate, up to three.
This tool does not search, and one call answers with one voice. Pass a candidate's
catalog_voice_id, repeat for the others so the user can compare, and let them say
which is theirs. language is a family (ko) or a locale (ko-kr).
Inputs: language (required), catalog_voice_id (required). Returns: model, locale, played, next_step, +4 more
read_service_guideRead service guideARead-onlyInspect
The official Onepin documentation pages that answer a question about the service.
What Onepin does, plans and credits, runs and downloads, the dictionary, templates, the API.
Answer only from what it returns. question is matched against English page descriptions,
so pass a short English restatement of theirs.
Inputs: question (required). Returns: pages, next_step
run_workflowRun workflowADestructiveInspect
Start a run. Spends credits, so the user must approve the cost first.
Needs estimate_run's run_token and the SAME script_text. Without confirm=true it starts nothing and returns the price to show the user; call again with confirm=true once they agree. Returns run_id and status; then poll get_run_status.
Inputs: run_token (required), workflow_id (required), confirm, script_text, +3 more. Returns: run_id, status, created, next_step, +10 more
save_dictionary_entrySave dictionary entryADestructiveIdempotentInspect
Add a dictionary entry, or edit one by entry_id: a word, its locale, its IPA and a category.
Store only an IPA the user chose. Fields you omit on an edit stay as they are; clear erases one.
Inputs: ipa, type, word, clear, +3 more. Returns: ipa, type, word, locale, +4 more
scan_script_for_dictionary_entriesScan a script for dictionary entriesARead-onlyInspect
Which of the workspace's dictionary entries this script actually says, with how often.
The list is sized by the script, not by the dictionary, so it stays short however many entries exist. A run applies all of them by default.
Inputs: locale (required), script_text (required). Returns: matches, next_step, truncated, applied_by_default
select_workspaceSelect workspaceAIdempotentInspect
Choose which workspace the other tools act in. Remembered across sessions.
Inputs: workspace_id (required). Returns: selected, workspaces
set_workflow_voiceSet workflow voiceADestructiveIdempotentInspect
Pin a catalog voice onto one locale of a workflow's generator node.
An ambiguous locale, model or generator returns candidates instead of guessing — show them and call again with that argument set. Last write wins. Existing runs are unchanged; re-running re-charges credits.
Inputs: workflow_id (required), catalog_voice_id (required), model, locale. Returns: note, model, locale, voice_name, +1 more
suggest_pronunciationsSuggest pronunciationsARead-onlyInspect
IPA candidates for a word, for the user to choose between. They differ on purpose and are not ranked.
Saves nothing. Supported: en-us, en-gb, de-de, fr-fr, es-es, es-mx, it-it, pt-pt, pt-br, ko-kr, ja-jp, zh-cn.
Inputs: word (required), locale (required). Returns: next_step, candidates
update_workflowUpdate workflowADestructiveIdempotentInspect
Replace an existing workflow's graph with an edited definition.
Read the current graph with get_workflow(include_definition=true), edit it, then validate_workflow_definition for the token. Call without confirm first: nothing is written and you get what it replaces plus an edit_token, which confirm=true requires.
Inputs: definition (required), workflow_id (required), validation_token (required), confirm, +1 more. Returns: name, note, updated, replaces, +8 more
upload_audio_fileUpload an audio fileAInspect
Get an audio file (.wav, .mp3 or .flac) into Onepin for a source_audio node to use.
Call it with no method first: nothing is uploaded and it returns the two ways to do this for the USER to choose between. Then call again with the one they picked.
Inputs: method, filename, workflow_id. Returns: Four disjoint branches, which is why every key is optional even for its own bra…
upload_script_fileUpload a script fileAInspect
Get a script file (.txt or .pdf) into Onepin for a workflow to read.
Call it with no method first: nothing is uploaded and it returns the two ways to do this for the USER to choose between. Then call again with the one they picked.
Inputs: method, content, filename, workflow_id. Returns: Four disjoint branches, which is why every key is optional even for its own bra…
validate_workflow_definitionValidate workflow definitionARead-onlyInspect
Check a workflow definition against every gate a run start applies. Saves nothing.
On success returns a validation_token create_workflow requires — it expires in 10
minutes and is bound to these exact bytes, so validate what you will create.
On failure each error carries a workaround naming the field to change.
Inputs: definition (required). Returns: valid, errors, strike, blocked, +5 more
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Related MCP Connectors
Generate AI voices, text-to-speech, subtitles, transcription, and video localization.
Natural AI voiceovers: search 500+ voices, generate speech, design and save custom voices.
Text to speech and transcription for any AI model: MP3 voiceovers, audio and YouTube to text.
Voice agents that book into the software your clients actually run, built and audited from chat.
Related MCP Servers
- FlicenseNot gradedqualityBmaintenanceEnables AI agents to generate narration in a single fixed, pre-approved voice, returning audio with per-line timing, hashes, and automatic quality checks on content, speaker identity, and pace, all locally and reproducibly.-
- AlicenseAqualityBmaintenanceEnables AI agents to turn scripts into narration audio with matching SRT/VTT captions in one tool call, plus voice cloning from a reference clip and transcription of local audio or video into subtitles.7MIT
- AlicenseAqualityBmaintenanceEnables AI agents to convert any-length text to speech in saved or cloned voices, generate SRT/VTT captions from local media, and produce per-scene narration with exact durations and frame timings for video timelines.5MIT
- AlicenseAqualityDmaintenanceVoice Agent Factory for Vapi phone agents: vertical templates, self-testing agents, audio forensics, prompt versioning, CI regression gates.628 npmMIT
Glama MCP Gateway
Add one secure layer between your agents and this server.