Fovea

Fovea turns uploaded videos into schema-conforming JSON via web app, API, or MCP server for developers and agents.

Overview

Fovea is a video-to-JSON service that turns an uploaded video into structured data defined by your schema. You provide the shape you want returned, either as a JSON Schema or a sketch of one, and Fovea responds with JSON that conforms to it.

The product is built for workflows where video needs to become machine-readable output rather than a manual review task. It supports a web app for one-off use, a direct API endpoint for code, and an MCP server for agent-based tools such as Claude Code or Codex.

Core capabilities

Schema-first extraction

Submit a video and provide the structure you want back as a JSON Schema or a schema sketch. Fovea fills that shape rather than asking you to design a prompt or preset.

Validated responses

Every result is validated before it leaves the system, so the returned JSON conforms to the schema you supplied. If the shape is wrong, Fovea retries on its side.

Timestamp-aware fields

Any field in the schema can carry a timestamp, allowing outputs to point back to a moment in the video. This is useful when you need to review or jump to a specific segment.

Web, API, and MCP access

Use the product through a web app, a single POST to /v1/extract, or an MCP server at /v1/mcp for agent workflows. The same account and billing meter apply across surfaces.

Ephemeral video handling

The site states that videos are deleted after the job finishes and that submitted content is not used to train a model. That makes the workflow closer to a disposable extraction job than a persistent media store.

Practical use cases

Meeting note extraction
Convert meeting recordings into structured fields such as titles, participants, decisions, and action items so they can be indexed or sent into downstream systems.
Clip summarization with metadata
Pull key moments, timestamps, topics, and sentiment from a clip when you need machine-readable metadata for search, review, or follow-up.
Agent-based video workflows
Let an agent like Claude Code or Codex inspect video through MCP and decide what schema it needs for a task without building a separate video-processing pipeline.
Schema prototyping
Use the web app to test a schema against a sample video before wiring the same structure into code, which is helpful for validating output shape during setup.
Direct developer integration
Send a video directly from curl, Python, or TypeScript when you want a simple integration path without installing a dedicated SDK.

Pros and Cons

Pros

Supports a schema-first workflow instead of prompt engineering or manual post-processing.
Returns validated JSON that conforms to the structure you request.
Offers three access paths: web app, API, and MCP server.
Can include timestamps in fields, which helps trace extracted data back to the original video.
States that uploaded videos are deleted after processing and are not used for model training.

Cons

The current launch limit is videos up to ten minutes, so longer source files are not supported yet.
The site guarantees schema conformance, not that every extracted judgment will be right, so users still need to review outputs when accuracy matters.
Pricing details are still partial on the site, and the pricing page itself currently returns a 404.

FAQ

How do I use Fovea?

You create a video-to-JSON request by sending a video and a JSON Schema, or a schema sketch, to the web app, API, or MCP server. The response is returned on the same request, so there is no separate job queue to manage.

How long can a video be?

Fovea says it supports videos up to ten minutes at launch. The site also says this is a current extraction limit rather than a model limitation, and that it is the first limit they plan to raise.

What model does Fovea use?

Fovea runs on Google's Gemini multimodal models today. The company says the layer on top is model-independent, so the underlying model can change without requiring schema changes.

What happens to uploaded videos?

Fovea deletes your video once the job finishes and says that nothing you send is used to train a model.

Does Fovea guarantee the content will be correct?

Fovea guarantees the returned shape and validation against your schema, not the correctness of every judgment inside the output. The site recommends adding timestamps to fields when you want to spot-check results against the source video.

Quick Facts

Category: Developer Tool
Primary workflow: Video upload to structured JSON output
Access methods: Web app, API, MCP
API endpoint: POST /v1/extract
MCP endpoint: https://api.fovea.run/v1/mcp
Video limit at launch: Up to 10 minutes

Fovea Alternatives

ByteAsk

ByteAsk is a terminal-first AI coding agent for C and C++ that edits repos and verifies changes with compilers, debuggers, sanitizers, and tests.

MeetStream

MeetStream is a meeting bot API for Zoom, Google Meet, Microsoft Teams, and Webex. Record, stream, and analyze meetings with usage-based pricing and $5 free credit.

CreateOS Sandbox

CreateOS Sandbox is an isolated compute environment for running code and agent workloads in Firecracker micro-VMs with private networking and SDK, CLI, or MCP control.

GitBot

GitBot is a local browser workspace for reusable coding bots powered by Claude Code, Codex, or OpenCode, with persistent threads and configurable permissions.

hob

hob is an independent workspace for coding agents, with local control over sessions, terminals, history, routing, and follow-up work.

Grok 4.7

Grok 4.7 is an xAI model for coding and knowledge work, available via Grok Build, Cursor, the Grok API and more, from $2 per million input tokens.