Nanonets — Full Profile (2026-09)

Nanonets, Inc. is headquartered in San Francisco with additional operations in Bengaluru. Founded in 2017 by Sarthak Jain and Prathamesh Juvatkar, the company went through Y Combinator Winter 2017 batch. It closed a $29 million Series B funding round in April 2024 led by Accellead, Elevation and Y Combinator, bringing total funding to $40.5 million. It remains an independent, VC‑backed private business.

Website

Field	Value
Official Website	https://nanonets.com · pricing: https://nanonets.com/pricing · docs/API: https://nanonets.com/api · MCP: no first-party `mcp.nanonets.com`; Composio gateway `https://mcp.composio.dev/nano_nets` (OAuth/API-key, 11 tools)
Company	Nanonets, Inc. has offices in San Francisco, California and Bengaluru. Founded in 2017 by Sarthak Jain and Prathamesh Juvatkar, IIT‑GN alumni and former Cubeit team members, the company joined the Y Combinator Winter 2017 batch. It closed a $29 million Series B financing round in April 2024 backed by Accel, Elevation and Y Combinator, bringing total capital raised to $40.5 million. The firm remains an independent, venture‑backed private business — not acquired, not publicly traded, not PE‑owned, and not bootstrapped. It serves over 1,000 enterprise customers, with 35 % of Fortune 500 companies among its clients per vendor claims. Compliance certifications include SOC 2 Type II, ISO 27001, GDPR and HIPAA, and its infrastructure runs on AWS and GCP. Nanonets maintains developer mindshare through open‑source assets: the MIT‑licensed DocStrange library and the Nanonets‑OCR‑s Hugging Face model; these open‑source offerings are separate from its commercial SaaS platform.
Status	Active, VC-independent, not shut down
Category	ai-productivity-data
Pipeline Stage	none

Title

Field	Value
Product	Nanonets
Descriptive Title	Nanonets — Intelligent Document Processing (IDP) + Agentic Workflow Platform: Template-free OCR-3 (VLM, #1 on vendor IDP leaderboard Apr 2026) + 300+ prebuilt models (invoice/PO/receipt/ID/bank-stmt/contract) + No-Code Workflow Builder + Finance Agents (AP/close/reconciliation) + Context Graph “Trail” + REST v2 + Webhooks + Composio MCP gateway (Claude/Cursor/Codex, 11 tools, 3rd-party) + NetSuite/SAP/Oracle/QuickBooks/Xero/D365 2-way sync; Starter $0/$ 50 credits then $100/100 credits, block$ 0.02/ $0.10/$ 0.30, Growth/Enterprise custom (2017 Jain+Juvatkar SF+Bengaluru Accel B)
One-line Positioning	Finance team points Nanonets at a shared inbox/PDF drop/S3 → OCR-3 extracts tables+fields template-free → workflow routes by confidence/amount/entity to approve/post-to-NetSuite/email-vendor → Agent layer closes AP matches 70-90% straight-through, Trail context-graph resolves edge cases (return policy vs ERP vs call-notes) → REST v2/webhook pushes to warehouse, Composio MCP lets Claude Code upload training images + pull predictions (not first-party MCP); billing is per-block-run metered (simple $0.02 / std-AI$ 0.10 / complex-AI $0.30), no per-seat platform fee, credits never expire; reverse Motion/Todoist per-seat, reverse Qualtrics per-interaction.

Features

Field	Value
OCR / Extraction	OCR-3 proprietary VLM (Apr 2026, vendor claims #1 vs GPT-5/Gemini/Claude on IDP leaderboard), template-free, 40+ languages, tables/line-items/confidence scores, `/parse /extract /split /chunk /vqa` endpoints; 300+ pretrained models (invoice/PO/receipt/ID/bank-statement/contract/BOL/customs); up to 99% field accuracy claimed (64.5 olmOCR-bench)
Workflow Builder	No-code drag-drop, conditional branches, human-in-the-loop approval nodes, email/cloud-intake (Gmail/Outlook/Drive/Dropbox/SharePoint), schedule/trigger, custom Python blocks, generative-AI blocks
Agent Layer (2025-26)	AP automation agent (invoice→3-way match→post ERP), month-end close, contract review, freight audit, KYC, trade-deduction; vendor cites 94% complex-process closure without human, 70-90% STP on mature AP; Trail context-graph ties policy/ERP/call-notes for edge-case reasoning
Integrations (native 2-way)	NetSuite (OneWorld), QuickBooks Online (Desktop via IIF/CSV), Xero, Sage Intacct, SAP S4HANA/B1 (Enterprise), Oracle Fusion (Enterprise), Salesforce, HubSpot, Microsoft Dynamics 365, Snowflake/BigQuery (export)
API	REST v2 (app.nanonets.com/api/v2/OCR/Model/{id}, Basic Auth API-key, async jobs, webhooks on prediction-complete/model-trained), SDKs Python/Node/Java/Go, no GraphQL
MCP	3rd-party via Composio `mcp.composio.dev/nano_nets` — 11 tools (create/list/delete OCR models, extract, upload training images, fetch predictions), OAuth or API-key, Claude/Cursor/Codex/Claude Agent SDK; no first-party Nanonets-hosted MCP server as of 2026-09 (distinct from Klaviyo/Feedly first-party GA, aligns Smartlead/Lokalise pattern where MCP exists but partner-mediated)
Security	SOC 2 Type II, ISO 27001, GDPR, HIPAA (Enterprise), SAML/SSO/SCIM/OIDC, RBAC, audit logs, SIEM, private-cloud/on-prem Enterprise, US/EU/APAC residency

Pricing (2026-09 USD, block-run metered, no seat fee)

Field	Value
Starter	$0 —$ 50 free credits (no card), then $100/mo for 100 credits (=$ 1/credit equiv), 3 users, API access, email/cloud connectors, community support; credits never expire, no platform fee
Block Rates (all paid)	Simple op $0.02/run (format/routing/export) · Standard AI $ 0.10/run (classify/validate) · Complex AI $0.30/run (data extraction/generative); typical invoice = 4-6 blocks → <$ 2/doc end-to-end
Growth	Custom quote — volume pricing, shared credits, up to 40% volume discount, classification+genAI blocks, ERP/database integrations, analytics, up to 40 users, team credit pooling
Enterprise	Custom — SAML/SSO/SCIM, HIPAA, private cloud/on-prem, US/EU/APAC residency, SAP/Oracle connectors, SLAs, audit/SIEM, white-label, dedicated CSM; floor often ~ $20K/yr historically, AP-scale deployments$ 50K-200K/yr
Legacy Page Note	Older nanonets.com/pricing variant shows “ $0.30/page Starter,$ 999/workflow/mo Pro” — superseded by credit/block model in 2026; print current block model as canonical, legacy per-page as historical
Type	block-run-metered-not-per-seat-not-per-doc-flat-not-per-call; starting_paid=$100/100-credits; no per-seat (reverse Motion/Todoist)
Note	Credit math: 1 credit ≠ 1 block; $100=100 credits, complex-AI block$ 0.30 ≈ 0.3 credit — finance teams must model runs/doc not seats

Reviews

Field	Value
G2	4.5/5 (~80 verified across “Nanonets” IDP profile, 2026) — primary B2B star; praise template-free extraction adapts to layout drift, invoice AP accuracy out-of-box, API-first dev ergonomics, Composio MCP lets Claude prototype; gripes block-pricing hard to forecast at variable volume, Enterprise connector quotes opaque (SAP/Oracle), QuickBooks Desktop only IIF/CSV not real-time, support slow pre-Enterprise, versioned model retrain UI clunky
Capterra	4.9/5 (~80 verified, G2-acquired Feb 2026 corpus separate) — auxiliary, vendor-value 4.8 ease 4.7 support 4.5; barbell rare, leans positive (IDP buyers review more willingly than ESP buyers)
Trustpilot	No substantive Nanonets brand page (B2B-only, no B2C surface) → `no-rated-consumer-profile`, excluded from B2B avg (aligns FFmpeg/MonkeyLearn-standalone/Smartlead pattern)
Analyst	Everest Group IDP PEAK Leader Apr 2026 (with ABBYY/Hyperscience/MS), Gartner IDP directory Dec 2025 (no MQ score) — tertiary
Editorial	ToolRadar 4.4, Infoseemedia Mar 2026 “ease of setup + API-first + invoice accuracy” praise, AI Productivity 2026 “48hrs/wk saved” vendor-cited
Praise	Kills template-maintenance tax (layout drift handled), OCR-3 tops vendor leaderboard, AP agent + NetSuite 2-way closes 70-90% STP, no per-seat = finance loves at scale, Composio MCP = devs prototype extraction in Claude Code afternoon, DocStrange/OCR-s open-source earns dev trust
Gripes	Block-run metering unpredictable for spiky volumes (invoice-heavy months surprise), SAP/Oracle connector only Enterprise + custom quote, QB Desktop second-class IIF/CSV, HIPAA/SSO gated to Enterprise, no first-party MCP (Composio dependency = another vendor in chain), vendor claims 35% F500 unverified

Review Sources Breakdown

Field	Value
G2.com	4.5/5 ~80 verified — primary B2B star, weight highest
Capterra.com	4.9/5 ~80 — auxiliary, G2-owned Feb 2026 corpus separate, IDP buyers lean positive
Trustpilot	none → `no-rated-consumer-profile` excluded
Everest/Gartner	Leader PEAK 2026 / directory listing — analyst not star
Editorial (ToolRadar/Infoseemedia/AI Productivity)	4.4 band + dev-ergonomics confirm
Conclusion	Prints “G2 4.5 primary, Capterra 4.9 aux, Trustpilot none excluded, Everest Leader 2026” — split-source rule holds; no TP barbell because B2B-only (unlike Klaviyo/Feedly/Motion)

Access

Field	Value
Web App	https://app.nanonets.com — model training UI, workflow builder, human-in-loop queue, Trail context graph viewer, dashboards; no desktop app
API	REST v2 (Basic Auth API-key, `/api/v2/OCR/Model/{id}`, async predict, webhooks `prediction_complete`/`model_trained`), SDKs Python/Node/Java/Go, batch upload, confidence thresholds
MCP	Composio 3rd-party gateway `https://mcp.composio.dev/nano_nets` — 11 tools, OAuth 2.1 or API-key, Claude/Cursor/Codex/Claude Agent SDK/CrewAI/LangChain; no first-party mcp.nanonets.com; aligns Smartlead/Lokalise “MCP exists but not vendor-hosted” subcluster, distinct from Klaviyo/Feedly first-party GA and from Motion/Todoist/Qualtrics none
Integrations	NetSuite/QBO/Xero/Sage/SAF/Oracle/Salesforce/D365 native 2-way; Gmail/Outlook/Drive/Dropbox/SharePoint intake; Zapier/Make; Snowflake/BigQuery export; Composio 1,000+ app bridge
Login	Google/Microsoft SSO; SAML/OIDC/SCIM Enterprise; workspace RBAC
Self-host	Private cloud / on-prem Enterprise only; cloud-default AWS+GCP, US/EU/APAC residency

Best For

Field	Value
Audience 1	Finance/AP teams (2-50 ppl) processing 1K-200K invoices/yr wanting 70-90% STP into NetSuite/QBO without template hell
Audience 2	Dev-heavy teams building custom extraction on proprietary forms (contracts/BOL/customs) via API + Composio MCP prototyping in Claude Code
Audience 3	Logistics/ops teams extracting POs/BOLs/lading from email scans into ERP
Audience 4	Regulated orgs (healthcare/insurance) needing HIPAA + on-prem Enterprise with audit/SIEM
Audience 5	CFO-tech evaluators comparing block-metering vs Docsumo/Rossum fixed-fee at mid-volume

Not For

Field	Value
Exclude 1	Sub-100-doc/mo solo users — $50 credit lasts but workflow overhead > manual keying (→ manual or free OCR)
Exclude 2	Buyers demanding first-party vendor-hosted MCP (Nanonets only via Composio 3rd-party; reverse Klaviyo/Feedly)
Exclude 3	Teams refusing block-run metering (want flat per-page or per-seat) — block math surprises at spike volumes
Exclude 4	QuickBooks Desktop real-time sync seekers (IIF/CSV only, not live 2-way)
Exclude 5	Pure OCR-without-workflow buyers (→ AWS Textract/Azure DI cheaper, no agent/ERP layer)
Exclude 6	Non-document unstructured data (audio/video) — text/PDF/image only

Competitors

Field	Value
ABBYY Vantage/FlexiCapture	Enterprise IDP incumbent, stronger SAP-deep docs, weaker no-code agent layer, pricier
Hyperscience	Mailroom-grade high-volume, ML-assisted, less dev-first API ease
Rossum	Invoice-focused IDP, fixed-fee predictable, narrower doc scope than Nanonets
Docsumo	SMB-mid invoice/loan docs, flat-fee friendlier, smaller model zoo
UiPath Document Understanding	RPA-bundled extraction, per-operation 2026 pricing, heavier platform lock-in
AWS Textract / Azure DI / Google Document AI	Raw OCR API, no workflow/agent/ERP, dev-self-assemble
Open-source (Unstract + DocStrange + Nanonets-OCR-s)	Self-host IDP, no SLA/UI/ERP connectors