Dewey
Ingest docs, search semantically, get cited AI answers
Last verified:
What is Dewey?
Dewey is a document backend for AI applications that turns uploaded files into searchable, cited answers. It accepts common document types such as PDFs, Word files, PowerPoint decks, web pages, and plain text, then handles conversion, section extraction, chunking, embedding, and retrieval automatically.
It is designed to provide real-time ingestion and hybrid retrieval, with section-aware search and grounded answers drawn from exact passages. The product emphasizes structure preservation, so headings, sections, and offsets remain available rather than being flattened into anonymous text chunks.
Dewey also supports hosted agents, letting users save an agent configuration with prompts, tools, and model settings, then run it from the dashboard, REST API, CLI, Python SDK, TypeScript SDK, or MCP-compatible assistants like Claude and Cursor. This makes it useful for teams that want document Q&A, research workflows, or agent orchestration without building the plumbing themselves.
The platform is aimed at developers, researchers, and teams that need reliable document search and cited research over internal or uploaded corpora. It is especially suited to people building AI apps, research tools, or assistant workflows that need citations, transparency, and flexible depth controls.
Dewey pricing
Pricing model: Freemium
Dewey offers a Free plan at $0 for prototyping and evaluation, with 1 project, 50 documents, 500 MB storage, 2,000 queries per month, 500 processing units per month, 25 research credits per month, and quick plus balanced research depth. Free projects are suspended after 90 days of inactivity. The Pro plan costs $49 per month and includes everything in Free, plus 5 projects, 2,000 documents, 10 GB storage, 50,000 queries per month, 20,000 processing units per month, 500 research credits per month, AI image and table captioning, near-duplicate detection, and email support. Pro also supports bringing your own API key, which unlocks deep and exhaustive research plus claim extraction and contradiction detection without Dewey charging research credits or generation fees. The Team plan costs $199 per month and includes everything in Pro, plus 20 projects, 50,000 documents, 100 GB storage, 250,000 queries per month, 250,000 processing units per month, 2,000 research credits per month, up to 20 team members, and priority email support. Enterprise pricing is custom and adds unlimited projects and seats, custom volumes, a private deployment option, custom embedding endpoints, SSO/SAML, SLA-backed uptime and support, and annual invoicing. Pro and Team overage rates are $0.02 per document, $1 per 1,000 queries, and $0.002 per processing unit. One processing unit equals one PDF page or one PPTX slide, and other formats count as one unit each.
Dewey pros
- Upload PDFs, Word docs, PowerPoint decks, web pages, and plain text
- Automatic conversion, chunking, embedding, and indexing
- Real-time ingestion pipeline
- Hybrid retrieval across documents
- Section-aware search with exact passages
- Cited answers linked to source sections
- Structured documents with preserved headings and offsets
- Four research depth levels
- Quick, balanced, deep, and exhaustive research modes
- Works with Claude and Cursor via MCP
- REST API for app integration
- CLI for terminal workflows
- Python SDK support
- TypeScript SDK support
- Hosted agents with saved configurations
- Full run transparency with tool-call traces
- No infrastructure to manage
- Free tier available
- No credit card required to start
- BYOK support for OpenAI, Anthropic, or Gemini
- AI image and table captioning on Pro+
- Near-duplicate detection
- Claim extraction on Pro+
- Contradiction detection on Pro+
Dewey cons
- Free tier limited to 1 project
- Free tier limited to 50 documents
- Free tier limited to 500 MB storage
- Free tier limited to 2,000 queries per month
- Free tier limited to 500 processing units per month
- Free projects suspend after 90 days of inactivity
- Deep and exhaustive research require BYOK
- Pro plan still has usage caps
- Team plan still has usage caps
- Overage charges apply on Pro and Team
- Research credits are metered on non-BYOK usage
- Processing units are metered for uploaded content
- Enterprise pricing is custom and not publicly listed
- Private deployment is only on Enterprise
- SSO/SAML is only on Enterprise
- Some advanced capabilities are tied to Pro+
- BYOK requires matching provider keys
- No unlimited free usage
Frequently asked questions about Dewey
What does Dewey actually do?
Dewey is a document backend for AI applications that lets you upload collections of documents and ask questions in plain English. It converts, structures, chunks, embeds, and retrieves content for you, then returns answers grounded in the exact source passages.
What file types can I upload?
Dewey accepts PDFs, Word documents, PowerPoint decks, web pages, and plain text. The site says it handles conversion automatically, so you do not need to preprocess files before upload.
How are answers grounded in the source documents?
Dewey returns cited answers drawn from the exact passages in your documents. It preserves document structure such as headings, sections, chunks, and offsets, which makes it possible to cite by section name instead of treating everything like a flat text dump.
What research depths are available?
Dewey offers four depth levels: quick, balanced, deep, and exhaustive. Quick is meant for factual lookups and targeted questions, balanced is the default for most Q&A, deep is for synthesis across larger corpora, and exhaustive is for complex analysis and full corpus traversal.
Do deep and exhaustive searches cost more?
On standard plans, deeper research uses more research credits because it is more expensive to serve. If you bring your own OpenAI, Anthropic, or Google Gemini key, Dewey says deep and exhaustive become quality settings rather than credit-based billing, and you pay your provider directly.
Can I use Dewey inside Claude or Cursor?
Yes. Dewey provides an MCP server so your document collections can be used inside Claude, Cursor, and any MCP-compatible AI assistant. The site positions this as a way to bring your knowledge base into the tools you already use.
What is included in the free plan?
The free plan includes 1 project, 50 documents, 500 MB storage, 2,000 queries per month, 500 processing units per month, 25 research credits per month, and quick plus balanced research depth. It does not require a credit card, but free projects are suspended after 90 days of inactivity.
What is included in Pro?
Pro costs $49 per month and adds 5 projects, 2,000 documents, 10 GB storage, 50,000 queries per month, 20,000 processing units per month, 500 research credits per month, AI image and table captioning, near-duplicate detection, and email support. It also supports BYOK for OpenAI, Anthropic, or Gemini keys, which unlocks deep and exhaustive research plus some advanced analysis features.
What is included in Team?
Team costs $199 per month and includes everything in Pro, plus 20 projects, 50,000 documents, 100 GB storage, 250,000 queries per month, 250,000 processing units per month, 2,000 research credits per month, up to 20 team members, and priority email support. It is positioned for small teams with multiple apps or environments.
How does hosted agent support work?
Dewey lets you create saved agents that bundle a system prompt, tool set, collection scope, and model into a reusable configuration. You can run those agents from the dashboard, REST API, CLI, Python SDK, TypeScript SDK, or scripts and CI, and the site says every run includes a full tool-call trace with inline citations and clickable source rows.