Schema-driven extraction
You define the shape of the output as a JSON Schema or a simpler schema sketch, and Fovea fills it with matching data instead of returning an unstructured summary.
Fovea
RevendiquerFovea turns uploaded video into structured JSON that matches a schema you supply, through a web app, REST API, or MCP server. It is aimed at workflows where teams or agents need video-derived data that is validated before use.
Video-to-JSON extraction for structured video workflows
Fovea is a video-to-JSON API and companion web app that accepts a video plus a schema and returns JSON matching that schema. The product is positioned for teams and agents that need structured data from video without building an extraction pipeline themselves.
The service can be used through the web interface, a REST endpoint, or an MCP server. Fovea says responses are validated against the supplied schema before they are returned, and that any field in the schema can carry a timestamp when the use case needs direct reference back to the source video.
Core capabilities
Validated responses
The service validates each response against the supplied schema before returning it, so the output is meant to be consumed by code without a separate conformance step.
Web, API, and MCP access
Fovea can be used in a browser, via a single POST to the REST endpoint, or through an MCP server for supported clients such as Claude Code and Codex.
Timestamped fields
Any field in the schema can carry a timestamp, allowing extracted items to point back to specific moments in the source video.
Synchronous return path
The documentation says the extraction comes back on the same request, so there is no polling loop or webhook setup for the standard workflow.
Account and data handling controls
The site states that videos are deleted after the job finishes and that uploaded content is not used to train a model.
Practical ways to use Fovea
Meeting and standup capture
Extract decisions, action items, participants, or a meeting summary from recorded meetings so the result can flow into notes, task systems, or internal tooling.
Agent-driven video analysis
Give an MCP client access to the extract tool so an agent can choose the schema it needs and pull structured facts from video on its own.
Application integration with a single endpoint
Send an uploaded file and a schema to the REST API when you want a lightweight integration path without an SDK or separate processing infrastructure.
Timestamped review workflows
Use timestamp fields when the output needs to support spot-checking or jumping directly to relevant moments in the video.
Schema-first experimentation
Try a shape in the web app before wiring it into code, which is useful when you are still deciding which fields the workflow should collect.
Pros and Cons
Pros
- Supports multiple access paths: web app, REST API, and MCP.
- Returns validated JSON that matches the supplied schema.
- Allows timestamped output fields for video-linked references.
- No SDK is required for the main API workflow.
- The site states that videos are deleted after processing and not used for model training.
Cons
- At launch, videos are limited to ten minutes.
- The docs say the system guarantees shape, not judgement, so extraction quality still needs review for important fields.
- Pricing is structured around video minutes, but the public pricing page was unavailable in the provided source set, so some plan details are only partially documented.
FAQ
Does Fovea require a full JSON Schema?
No. The site says you can provide either a full JSON Schema or a looser sketch of the shape you want back.
How do I integrate it with an agent?
Fovea exposes an MCP server that can be added in one line to supported clients such as Claude Code or Codex, and the agent then uses the extract tool with its own schema.
Does the API return results synchronously?
Yes. The documentation says the extraction comes back on the same request, so there is no polling flow in the normal API path.
What are the main limits on video length?
The docs state that Fovea accepts videos up to ten minutes at launch.
What happens to uploaded videos?
The site says videos are deleted once the job finishes, and that uploaded content is not used to train a model.
Quick Facts
- Category
- Developer Tool
- Primary interface
- Web app, REST API, MCP server
- Output
- Validated JSON matching a supplied schema
- Supported clients
- Claude Code, Codex, any MCP client
- Video limit at launch
- Up to 10 minutes per video
- Data handling
- Videos deleted after processing; not used for training
Analyses de Fovea
Fovea· Visites mensuelles 0
Les données de trafic sont fournies à titre indicatif uniquement.
- Visites mensuelles
- 0
- Classement mondial
- -
- Classement de catégorie
- -
- Taux de rebond
- 0.00%
- Durée moyenne de visite
- 00:00
- Pages par visite
- 0.00
Tendances du trafic
| mai | 0 |
|---|---|
| juin | 0 |
| juil. | 0 |
Sources de trafic
Les analyses de trafic ne sont pas encore disponibles.
Principales régions
Les analyses de trafic ne sont pas encore disponibles.