Captain - Multi-Modal RAG API for Video, Docs, and Images

TL;DR

TL;DR: Captain is a managed API for multi-modal RAG — index any file type (video, images, PDFs, audio) with two API calls and query across all of them at once. Backed by Y Combinator W26.

Source and Accuracy Notes

⚠️ This section is MANDATORY. All links must be verified from actual source, not guessed.

• Project page: runcaptain.com — verified Jul 2026
• Documentation: docs.runcaptain.com — verified Jul 2026
• HN launch thread: news.ycombinator.com/item?id=47366011 — 57 points, Jul 2026
• Pricing page: runcaptain.com/pricing — verified Jul 2026
• License: Proprietary SaaS (no open-source license)

What Is Captain?

Captain is a fully managed API for deterministic file search over unstructured data. Rather than stitching together separate pipelines for video, images, PDFs, and audio, Captain exposes two endpoints that handle the entire multi-modal RAG workflow:

1. Index — ingest files from cloud storage
2. Query — search across all indexed content in one call

The product is built by Captain Technologies, Inc. and backed by Y Combinator’s W26 cohort. It positions itself as a “great agents have great search” solution — the core bet being that retrieval quality determines downstream agent performance.

Core API Pattern

The demo on the landing page shows the canonical usage:

POST /v2/collections/{collection_id}/query

You send a query string, and Captain returns ranked results across all indexed modalities — video frames, document pages, images, audio segments — with relevance scores.

From the landing page, the product benchmarks at 95% accuracy across video, documents, and images on the MRAG-Bench benchmark (ICLR).

Supported File Types and Modalities

Based on the live product demo and documentation sidebar, Captain handles:

• Video — with frame-level timestamps and relevance scores
• Images / Handwriting — scanned documents, handwritten notes
• Documents (PDF) — with OCR and page-level citation
• Multi-lingual — Tagalog and other non-English documents shown in the demo
• Audio / Music — with transcript segments

This is more comprehensive than most RAG APIs, which typically handle one or two modalities well.

Integrations

Captain connects directly to cloud and enterprise storage:

• AWS S3
• Google Cloud Storage (GCS)
• Azure Blob Storage
• Dropbox
• SharePoint
• Google Drive
• Confluence
• Slack
• Gmail
• Notion

The Enterprise tier adds BYOC (On-Premise) deployment for teams that cannot send data to external APIs.

Pricing

Captain uses a credit-based model. From the pricing page (verified Jul 2026):

| Plan | Price | Credits/mo | |---|---|---| | Free | $0/mo | 3,000 | | Startup | $1,600/mo | 83,000 | | Enterprise | Custom | Unlimited |

Overage pricing is $0.015 per additional credit. The Free tier is limited to 3 collections; Startup and Enterprise tiers support unlimited collections.

Security

The product markets SOC 2 Type II certification (verified on the landing page footer). For enterprise customers, BYOC (bring-your-own-cloud) deployment is available on the Enterprise plan.

How Captain Compares to Other RAG Tools

From HN discussion on the launch thread, several alternatives were mentioned in comparison questions:

• qmd — Q&A over documentation as a service
• Danswer — open-source AI search over private data (YC W24)
• Greptile — RAG on codebases (YC W24)

Captain’s differentiator is its breadth of modality support (video, audio, handwriting in addition to text) in a single API call, and its focus on retrieval accuracy as a primary metric rather than just connecting to LLM context windows.

FAQ

Q: Is Captain open source? A: No. Captain is proprietary SaaS. There is no self-hosted or open-source version. Enterprise plans offer BYOC (bring-your-own-cloud) deployment, but the software itself is not open source.

Q: What happens to my uploaded files? A: Files are processed on Captain’s infrastructure. The Enterprise tier’s BYOC option lets you keep data within your own cloud environment. For the Free and Startup tiers, data is processed by Captain’s managed infrastructure.

Q: How accurate is the retrieval? A: The product claims 95% accuracy on the MRAG-Bench benchmark (ICLR) across video, documents, and images. Actual performance will vary by file type and query complexity.

Q: What is a “credit” in Captain’s pricing? A: Credits are consumed per API operation. The exact credit cost per query or indexing action is not publicly documented beyond the per-credit overage price ($0.015/credit).

Q: Does Captain support self-hosted deployment? A: Only on the Enterprise plan via BYOC. There is no public self-hosted option for Free or Startup tiers.

Conclusion

Captain solves the multi-modal RAG problem by making it a two-API-call operation: index from your existing cloud storage, then query across everything. The breadth of supported modalities — video, audio, handwriting, multi-lingual PDFs — is wider than most alternatives, and the SOC 2 Type II certification addresses the compliance concern that typically blocks RAG adoption in enterprise settings.

The main caveats: it is SaaS-only (no self-hosted option unless you’re on Enterprise), and the pricing at $1,600/mo for the Startup tier is oriented toward teams with real traffic. The Free tier with 3,000 credits and 3 collections is sufficient for evaluation.

If you’re building agents that need to reason over mixed media — meeting recordings, scanned documents, multilingual knowledge bases — Captain is worth evaluating. The single-query multi-modal retrieval is a genuine time-saver versus stitching together separate pipelines.

Source: runcaptain.com | HN: news.ycombinator.com/item?id=47366011