grep, but for video.
Ask a question. Search every frame and transcript. Give your AI agents the exact clips and timestamps they need—across local drives or S3.
$ vidgrep search ./launches \
--query "customer mentions slow exports"
Found 2 moments in 1.8s
product-interview.mp4
00:14:08 → 00:14:32 0.96
sales-call-042.mp4
00:31:17 → 00:31:49 0.91Search without scrubbing
Your video library,
finally queryable.
One search layer for spoken words and visual meaning, designed to fit directly into automated workflows.
Natural-language search
Search by meaning, actions, objects, or dialogue—not just filenames and metadata.
Regex transcript grep
Drop down to exact patterns with familiar regex search across every indexed transcript.
Timestamped clips
Get precise start and end timestamps, frame ranges, confidence scores, and ready-to-use clips.
S3 + local indexing
Point VidGrep at a folder or bucket. Incremental indexing keeps large libraries current.
Agent-native MCP server
Let Claude, Codex, and custom agents discover and call video search as a native tool.
REST API + SDK
Build with a clean API, typed responses, webhooks, and ergonomic JavaScript and Python SDKs.
How it works
From archive to answer
in three commands.
- 01
Index
Connect a local directory or S3 bucket. VidGrep transcribes, embeds, and tracks frame-level metadata.
- 02
Query
Use natural language, regex, the REST API, CLI, SDK, or MCP from inside your agent workflow.
- 03
Get clips
Receive ranked moments with timestamps, frame ranges, source paths, and clipped media ready for the next step.
Simple pricing
Search more. Scrub less.
Every plan includes semantic video search, timestamped results, and secure local or S3 indexing.
Starter
For solo builders shipping their first video-aware agent.
$19 / month
- 100 indexed hours
- 10,000 searches / month
- Local + S3 sources
- CLI and REST API
Team
For product teams putting video search into production.
$49 / month
- 500 indexed hours
- 50,000 searches / month
- MCP server + SDKs
- Priority indexing
Business
For video-heavy workflows that need room to scale.
$99 / month
- 2,000 indexed hours
- 200,000 searches / month
- Team access controls
- Priority support
From the terminal
Built for people
who build agents.
“We replaced a brittle pile of transcription scripts with one API call. Our agent now finds the right customer moment before I can open the file.”
“The frame ranges are the killer feature. VidGrep gives our editing agent exactly enough context to cut a usable clip on the first pass.”
“MCP made integration almost suspiciously easy. By lunch, our research agent could search three years of recorded demos.”
Does VidGrep upload my local videos?
No. Local indexing can stay local; only the generated index and query data move according to the deployment mode you choose. S3 sources are read with your scoped credentials.
What kinds of video can I search?
VidGrep works with common video formats and searches both speech and visual content, including objects, actions, scenes, and on-screen context.
Can an AI agent call VidGrep directly?
Yes. Use the bundled MCP server for native tool access, or call the REST API and typed SDKs from any agent framework.
How precise are the results?
Every result includes start and end timestamps, frame ranges, a confidence score, and source metadata. You can tune clip padding for your workflow.
Can I change or cancel my plan?
Yes. Plans are month-to-month. You can upgrade, downgrade, or cancel from your billing portal at any time.
Stop watching. Start finding.
Give your agents eyes for video.
Index your first library today and turn hours of footage into answers measured in seconds.