QC, conform and AI video, image and sound generation for film. Priced before every run.
Scored across 14 tools
The set is largely cleanly separated: runs (start/get/list/cancel/estimate), uploads (upload_file/complete_upload), and discovery (list_models/tools/voices/workflows). The only real soft spot is chat_completion versus start_run with a text model, which the descriptions acknowledge but still leave as two entry points for starting generation.
Almost everything follows a strict verb_noun pattern (start_run, get_run, list_runs, cancel_run, estimate_run, upload_file, complete_upload, create_topup_link, get_balance). The lone deviation is chat_completion, which is noun_action rather than verb_noun, but the set remains highly readable.
Fourteen tools sit squarely in the well-scoped 3-15 range, and each covers a genuine lifecycle step rather than being redundant. Nothing feels padded.
Full run lifecycle (estimate/start/poll/cancel/list), upload (create/complete), model/tool/voice/workflow discovery, balance, and billing top-up are all covered. Minor gaps exist: no file listing or deletion, no bulk cleanup of uploads, and no dedicated usage/history endpoint beyond list_runs.
Stops a queued or running run. What it held is refunded. A finished run is returned unchanged.
Inputs: run_id (required).
One chat completion on a text model: give the model (a text model id from list_models, modality text, or auto) and the messages; the answer is the model's message and the usage, with the credits charged. The most the request may cost is held first and the rest refunded. max_tokens bounds the answer and the hold: name it. Prices are per model in list_models. Text models are zero data retention except where marked: a model whose data_retention is not none keeps requests as that field says (for example 30 days by the model maker), and auto never picks one.
Inputs: model (required), messages (required), max_tokens, temperature, +1 more.
Finishes an upload after the bytes are sent: checks them and makes the file ready for runs. For a multipart upload, give the parts with the ETag each PUT returned.
Inputs: file_id (required), parts.
Makes a payment link for a credit pack. Give the link to a person: they pay on the payment page, and the credits land when the payment succeeds. The link expires after 24 hours.
Inputs: pack (required), promo_code, idempotency_key.
Prices a run without starting it: the most it may cost, in credits. Give one of tool, model and workflow, and the inputs. Nothing is written and nothing is charged.
Inputs: tool, model, inputs, workflow.
The workspace's credit balance (1 credit = 1 US cent).
Inputs: no named parameters.
Reads one run: status (queued, running, succeeded, failed, cancelled), outputs with signed links, the report of a QC tool, credits charged, and the error of a failed run. Poll it after start_run.
Inputs: run_id (required).
Lists the generation models by public id, with each model's inputs and its price per unit in credits at this workspace's rate. app_name is the name the Wazza app shows (for example Gemini 3.1 is nano-banana-2): when a person names a model, match it there, then pass the id. Run one with start_run and model. A text model (modality text, or auto) is used with chat_completion instead, where that tool is listed.
Inputs: no named parameters.
Lists runs started through the API in this workspace (or this key's project), newest first, a page at a time. Pass next_cursor back as cursor for the next page.
Inputs: tool, limit, model, cursor, +4 more.
Lists the tools: id, what each does, its inputs (name, type, required, allowed values, bounds), its price unit, and whether it can run through this server (available).
Inputs: no named parameters.
Lists the voices this workspace may use with the voice tool. cloned marks a copy of a real person's voice: ask before you use one.
Inputs: no named parameters.
Lists the workflows: runs of several steps that need no choice from a person once started (deliver, unslop, shorts). Each workflow's id, what it does, its inputs and its steps. Run one with start_run and workflow; price it first with estimate_run.
Inputs: no named parameters.
Starts a run of a tool, a model or a workflow and returns AT ONCE with the run (id, status queued or running). Runs take seconds to minutes: call get_run with the run id until status is succeeded, failed or cancelled, then read its outputs (signed links) and report. A workflow run also lists its steps, each with its own status and credits. The credits are held when the run starts and settled when it ends. Give one of tool, model and workflow, and the inputs; files are https URLs or file ids from upload_file.
Inputs: tool, model, inputs, workflow, +2 more.
Creates an upload for an input file and returns its file id and where to send the bytes: PUT them to upload.url with the headers it lists (or, for a large file, each part to its URL), then call complete_upload. Name the file in a run's inputs by its file id.
Inputs: name (required), size_bytes (required), content_type (required), sha256, +1 more.
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
AI film lab for filmmakers: generate and review images/clips with input provenance, cost preflight
AI film studio: shoot video, images and sound, review takes, audit scripts, quote before you shoot.
AI video production, idea to final cut: generate, edit and assemble films, ads and social video.
Film production: projects, shots, scripts, and picture and video jobs priced before they run.
Add one secure layer between your agents and this server.
