Skip to main content

Pricing

Pay for what you use, not what you build

Every plan includes the full ingestion pipeline, hybrid retrieval, section-aware search, and a complete REST API. No infrastructure to manage. Start free, no credit card required.

Free

$0

For prototyping and evaluation

Get started
  • 1 project
  • 500 MB storage
  • 2,000 queries / month
  • 500 processing units / month
  • 25 research credits / month
  • quick + balanced research depth

Free projects suspended after 90 days of inactivity.

Most popular

Pro

$49/ month

For developers who need exhaustive depth and full model flexibility

Start building
  • Everything in Free, plus:
  • 5 projects
  • 10 GB storage
  • 50,000 queries / month
  • 3,500 processing units / month
  • 500 research credits / month
  • AI image & table captioning
  • Near-duplicate detection
  • Email support

Bring your own API key unlocks

  • • Deep & exhaustive research depth
  • • Claim extraction
  • • Contradiction detection

Supply an OpenAI, Anthropic, or Gemini key. You pay your provider directly — no research credits, no processing units.

Team

$199/ month

For small teams with multiple apps or environments

Start building
  • Everything in Pro, plus:
  • 20 projects
  • 100 GB storage
  • 250,000 queries / month
  • 15,000 processing units / month
  • 2,000 research credits / month
  • Up to 20 team members
  • Priority email support

Enterprise

Custom

For orgs with compliance, scale, or isolation requirements

Contact us
  • Everything in Team, plus:
  • Unlimited projects & seats
  • Custom query & processing unit volume
  • Private deployment option
  • Custom embedding model endpoints
  • SSO / SAML
  • SLA-backed uptime & support
  • Invoiced annually

Pro & Team overage rates: $1 / 1k queries · $0.008 / processing unit. 1 unit = 1 PDF page or PPTX slide; other formats = 1 unit. Disable in your dashboard to hard-cap at plan limits.

How research credits work

Each research request consumes credits based on depth. Deeper analysis costs more because it's more expensive to serve. An exhaustive scan can traverse your entire corpus. BYOK bypasses credits entirely for research requests. When you bring your own OpenAI, Anthropic, or Google Gemini key, depth becomes a quality setting, not a billing one.

DepthCreditsBest forAvailability
quick0.5Factual lookups, targeted questions against known documentsAll plans
balanced1Default for most Q&A and agent workflowsAll plans
deepBYOKSynthesis across large or multi-document corporaPro+ with BYOK
exhaustiveBYOKMulti-hop reasoning, complex analysis, full corpus traversalPro+ with BYOK

Deep and exhaustive depths require a BYOK key matching the model provider (OpenAI, Anthropic, or Google Gemini). If a key is configured, it is used for all depths (including quick and balanced) and no credits are consumed. You pay your provider bill directly; Dewey charges nothing for the generation.

What we don't charge for

We deliberately keep certain dimensions off the bill to reduce cognitive overhead and remove friction.

API requests

Request volume is already captured by queries and documents. We never punish you for building an active app.

Egress / bandwidth

We don't charge for egress. Your data flows freely.

Collections

Collections are a logical grouping. We won't push you toward bad schema decisions to save money.

Real-time connections

WebSocket connections are core to the developer experience. Metering them would undermine the product.

Section summaries

AI section summaries are generated automatically at ingestion, cached permanently, and absorbed into plan pricing.

BYOK research tokens

When you bring your own key to /research, we absorb no generation cost, so we don't charge you for it.

Claim extraction

Claim extraction runs against your own API key and consumes no processing units. You pay your model provider directly; Dewey charges nothing for the run.

Contradiction detection runs

Contradiction detection uses your own API key and consumes no processing units. You pay your model provider directly; Dewey charges nothing for the run.

Near-duplicate detection

Deduplication reuses the chunk embeddings Dewey computed at ingest — no LLM calls, no processing units. Run it as often as you like; Dewey charges nothing for the run.

Have questions about plans, limits, or data? Read the FAQ →

Ready to build?

Start with the free tier. No credit card required. Upgrade when you need higher limits, deeper research, or a team.

Talk to us