Nanonets
Also known as: Nanonets OCR-3, Trail, docs.nanonets.com
Intelligent document processing and agent platform whose agents read documents, apply your rules, and post decisions into systems of record, closing finance and operations workflows without a person in the loop by default.
Nanonets turns document heavy business processes into AI agents, and is used by 35 percent of the Fortune 500 and more than 10,000 enterprises including Volkswagen, Procter & Gamble, Roche, Bayer, Mondelez, Schneider Electric and Ryanair. Its two primitives are agents and context graphs. An agent reads incoming documents in any format and with no per template setup, applies the customer's business rules, decides what to do, escalates when uncertain and writes structured results into the systems of record. Agents run multi step workflows without a human in the middle by default.
Trail, the context graph, is where the rules live. Policies, standard operating procedures and runbooks are uploaded and parsed into a machine readable graph in which every constraint is a node and every dependency an edge. Each rule is traceable to the page it came from and reviewed by a person before it goes live.
Trail also captures the tribal knowledge that never gets written down, extracting rules from Slack, Zoom, Teams and email, and logs every past decision as a precedent that later cases are matched against. Confidence scores decide when an agent acts alone and when it asks, corrections update the rulebook so autonomy rises over time, and the platform surfaces conflicts in the rules and proposes fixes for approval.
Extraction runs on Nanonets OCR-3, the company's own model, which Nanonets bills as first on the IDP leaderboard. Agents work across ERP and finance, procurement, order and revenue, supply chain, healthcare revenue cycle, storage, support and data platforms, covering accounts payable, reconciliation, order management, cash application, supplier onboarding and claims.
Teams can point the assistants they already run at Nanonets, including Claude, ChatGPT, Copilot, Bedrock, Vertex AI, SAP Joule and Agentforce, through an MCP server, native retrievers or a REST and GraphQL API, or use the pre built agents. Documentation is public at docs.nanonets.com. Every agent run is traceable and explainable and streams to the customer's SIEM.
Nanonets is SOC 2 Type II, ISO 27001 and GDPR compliant and signs a HIPAA business associate agreement. It offers SAML SSO, OIDC and SCIM, with RBAC across organizations, workspaces and agents. It deploys in a VPC, a single tenant cloud or on premises, with data residency pinned to US, EU or APAC regions and customer managed keys.
Nanonets was founded in 2017 in San Francisco and raised a $29 million Series B led by Accel. It is free to start with $50 in credits and bills on consumption above that.
Vendor details
Canonical URL
https://nanonets.com
Category
Enterprise operations agent
Company status
independent
Use cases & customers
In practice
A finance team hand keys three thousand invoices a month into Salesforce. Nanonets reads each one, matches it to the purchase order, routes exceptions to a reviewer, and posts clean data automatically.
An order arrives with a product code the agent is not sure about. Nanonets pauses below its confidence threshold and asks a named reviewer in Slack, Teams or email which code to use, and the correction updates the rulebook so the same case clears on its own next time.
A healthcare provider must keep patient records on its own infrastructure. Nanonets runs in the provider's virtual private cloud under a business associate agreement, extracting structured data from lab reports and forms without data ever leaving the network.
Agentic Index coverage score
11.0 / 14 capabilities · 79%
| Integrations & Tool Calling | Full |
|---|---|
|
The integration catalog spans nine categories and dozens of systems, including ERP and finance (SAP, Oracle, NetSuite, Dynamics 365, Workday, QuickBooks, Xero, Sage, Zoho, Odoo), procurement and contracts (Ariba, Coupa, Jaggaer, DocuSign, Ironclad, Icertis), order and revenue (Salesforce, HubSpot, Shopify, Stripe, PayPal, Adyen), supply chain (FedEx, UPS, DHL, Maersk, Flexport, project44), healthcare RCM (Epic, Oracle Health, Availity, Waystar, Optum), storage, support and data platforms. Agents act as well as read. They post decisions into systems of record, create sales orders, block a supplier whose bank details do not match the W-9, and adjudicate and pay claims. SourceNanonets, nanonets.comread 2026-09-07 |
|
| Workflow Orchestration | Full |
|
Agents run multi step workflows without a person in the loop by default. An agent reads incoming documents, applies the rules held in the context graph, decides what to do, escalates when uncertain and writes structured results into the systems of record, and agents are composed from a shared set of building blocks. Several agents take part in a single run with distinct scopes, such as an AP clerk agent, a regional manager agent and a business analyst agent in one freight audit run, and the claims example runs parse, rule evaluation, adjudication and payment in sequence. SourceNanonets, nanonets.com and docs.nanonets.comread 2026-09-07 |
|
| Knowledge Grounding & RAG | Full |
|
Trail, Nanonets' context graph, is built from the customer's own knowledge. Policy documents, SOPs and runbooks are imported and parsed into a machine readable graph where every constraint is a node and every dependency an edge, holding explicit, conditional, negative, precedence and implicit rules, each traceable to the source page it came from. It also absorbs tribal knowledge from Slack, Zoom, Teams and email as exception nodes and logs past decisions as precedents that later cases are matched against, with a case match score. The graph persists, is shared across every agent that needs it, picks the right rule for each situation, cites its sources and records which rule drove which decision. SourceNanonets, nanonets.com and docs.nanonets.comread 2026-09-07 |
|
| Human Oversight & Guardrails | Full |
|
Edge cases route to Slack, Teams or email for review, and confidence scores decide when an agent acts on its own and when it asks a person, as in an example where an agent pauses at 61 percent confidence to ask a named reviewer which product code to use. Escalation thresholds come from the customer's own policy and are encoded in the context graph, with the claims example routing by value to one or two reviewers with SLAs, and proposed self corrections go to a person to approve, review or reject before a rule changes. Every extracted rule is approved, edited or rejected by a person before agents run on it. SourceNanonets, nanonets.comread 2026-09-07 |
|
| Security, Identity & Governance | Full |
|
Nanonets holds SOC 2 Type II certification, ISO 27001 and GDPR, signs a HIPAA business associate agreement on enterprise plans and is independently audited every year. Enterprise controls cover SAML, OIDC and SCIM user provisioning with Okta, Azure AD and Google Workspace, full RBAC and fine grained access control across organizations, workspaces and agents, IP allow listing, AES-256 at rest and TLS 1.3 in transit with customer managed keys, and per team budgets and API quotas. The agent run view shows scoped, read only and configure level access for individual agents, backed by audit logs. SourceNanonets, nanonets.comread 2026-09-07 |
|
| Observability & Auditability | Full |
|
Every agent run is traceable, auditable and explainable, and a run shows a named run identifier, the invoice routed, the purchase order and load it matched, which rules fired and which did not, and which other agents took part with their access scope. Every agent run, approval and data access is recorded and streams to the customer's own SIEM by webhook, every agent decision traces to the document it read and the rule it applied, and each extracted field links back to its source for side by side citation. That record, kept by Nanonets itself, shows why an agent acted, not only that it did. SourceNanonets, nanonets.com and docs.nanonets.comread 2026-09-07 |
|
| Memory & State Persistence | Partial |
|
Two forms of persistence go beyond session state. Precedent memory logs every past decision with its full context and cites it as precedent, matching new cases against it with a similarity score, so a decision made once shapes later ones. Learned instruction state updates the agent's rulebook with each human correction, shown as a day 1 against day 30 rulebook where an exception the agent once escalated now clears. Neither is presented as a memory layer with a stated scope and lifetime, nothing says where it lives or how it can be deleted for one customer, and the rules themselves live in the context graph. SourceNanonets, nanonets.comread 2026-09-07 |
|
| Deployment & Data Residency | Full |
|
Customers can deploy privately in a VPC, in a single tenant cloud or on premises, on their own infrastructure and network policies, and data residency pins processing to US, EU or APAC regions, so customer data never leaves the customer's boundary. Bring your own key covers customer managed encryption keys. SourceNanonets, nanonets.comread 2026-09-07 |
|
| Prebuilt Agents, Templates & Packs | Full |
|
Customers pick and run pre built agents, including invoice and purchase order matching, bank statement and expense matching and contract analysis, and running one means picking a workflow and dropping in a document. Pre built agents are offered as the alternative to pointing a customer's own agents at the platform, and solution pages ship named agents for accounts payable, order management, logistics and healthcare. Each does its own job when selected. SourceNanonets, nanonets.com/agents and nanonets.comread 2026-09-07 |
|
| Triggers & Channel Coverage | Full |
|
Inbound email, including a named order intake mailbox, brings work to the agents, along with EDI transmissions, scans and faxes, PDF uploads, bank feeds and connected storage in Google Drive, SharePoint, OneDrive, Dropbox and Box, and reviewers are engaged over Slack, Teams or email with notifications on exceptions. Incoming documents and messages form an inbound queue, and feeds from ERP, payment and logistics platforms start work on events. SourceNanonets, nanonets.com and nanonets.com/agentsread 2026-09-07 |
|
| Model Flexibility & Routing | Not documented |
|
Nanonets runs its own proprietary extraction model, Nanonets OCR-3, and publishes open source models on Hugging Face, and no customer or admin selection of the model behind its agents is described. The Claude, ChatGPT, Copilot, Gemini, DeepSeek, Mistral and Hugging Face logos shown under connect your agent platform are outside assistants that call Nanonets through its MCP server or API, not a choice of the model behind its agents. SourceNanonets, nanonets.comread 2026-09-07 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
Customers can point the agents their teams already run at Nanonets through an MCP server, native retrievers, or a plain REST or GraphQL API, with Claude, ChatGPT, Copilot, Bedrock, Vertex AI, SAP Joule, Salesforce Agentforce, LangChain and LlamaIndex named as callers. The public documentation at docs.nanonets.com has an llms.txt index of every page, endpoints in OpenAPI form and markdown for any page, and open source models are published on Hugging Face. SourceNanonets, nanonets.com and docs.nanonets.comread 2026-09-07 |
|
| Testing, Debugging & Optimization | Partial |
|
After deployment, an optimization loop works on the agents themselves. Nanonets finds conflicts and undetected issues in the encoded rules, diagnoses them and proposes a fix for approval, for example noticing that Net-60 terms were granted twenty times by hand this quarter and recommending the rule change, and every human correction on an exception updates the instructions so autonomy rises with runs. No harness, scored test cases or gate before deployment is described, and no change is compared against a control, so the loop has no measured result. Auto-labeled samples and model retraining improve extraction accuracy but do not evaluate the agent. SourceNanonets, nanonets.comread 2026-09-07 |
|
| Browser & Computer Use | Not documented |
|
Agents reach outside systems through native integrations, retrievers and REST or GraphQL calls, and Nanonets delivers clean data to systems of record, so actions run through programmatic connections, not an interface an agent operates. Documents arriving as email attachments, EDI, scans, faxes and uploads are parsed, which is ingestion, not driving a screen. No browser, desktop or remote computer control is described. SourceNanonets, nanonets.comread 2026-09-07 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Pricing
Starter is free to start with $50 in credits and no card required, then $100 a month for 100 credits, paying only for what is used. Enterprise is quoted.
Consumption, charged per block run. Free signup credits come first, then usage based charging, with enterprise agreements above it.
What is public
The free entry point is published outright, with $50 in credits, no card required and self serve signup, then $100 a month for 100 credits on Starter. A pricing page and enterprise cost controls are published alongside it.
Billing mechanics
Consumption based. Free credits on signup, usage charging above them, and custom enterprise agreements for higher volumes with per team budgets and API quotas.
Additional watchouts
Cost tracks volume rather than seats, so exposure rises with automation. Per team budgets, quotas and cost dashboards limit it, and one customer case study cites zero overage charges.
Sales call required
Mixed (some tiers require a call)
Free / trial
Starter starts with $50 in free credits and no card required
Lowest paid plan
Starter at $100 per month for 100 credits, after the $50 in free credits
Support SLA / resale
Starter has community support, and Enterprise adds dedicated support and SLAs.
Missing data
Growth and Enterprise rates are quoted, not published.
Related vendors
- 11th Estate — Agentic platform whose AI engine scans markets, matches an…
- Abacum — FP&A platform for growing companies whose AI Modeler builds, edits…
- Adopt AI — AI native accounting and bookkeeping firm whose agents run…
- Aera Technology — Decision intelligence platform where an always on agent executes…
- Akro AI — On premise operational intelligence that automates document heavy…
- alfred_ — AI executive assistant that triages the inbox overnight and scores…
Alternatives to Nanonets
The closest documented capability profiles to Nanonets among enterprise operations agents tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Clio10.5 / 14A lighter documented profile than Nanonets
- Pigment10.5 / 14A lighter documented profile than Nanonets
- Workday10.5 / 14A lighter documented profile than Nanonets
- Anaplan10.0 / 14A lighter documented profile than Nanonets
- Glean12.0 / 14Adds documented Model Flexibility & Routing
- Leah AI11.0 / 14Adds documented Model Flexibility & Routing
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded