Tungsten Automation
Also known as: Kofax, Tungsten TotalAgility, TotalAgility, KTA, Kofax TotalAgility, Tungsten Capture, Kofax Capture, Tungsten RPA, InvoiceAgility, DocAI Studio
Document processing and workflow platform whose TotalAgility agents orchestrate across workflows, draw on knowledge bases, run on customers' chosen models, and pair with Tungsten's own screen-driving RPA, with FedRAMP High authority.
Tungsten Automation, formerly Kofax, builds TotalAgility, a platform that combines document processing, workflow automation, knowledge discovery and AI agents. Its agents coordinate with each other across enterprise workflows through multi-agent orchestration, answer questions from knowledge bases that load documents of any size for retrieval with source-backed answers, and let users query case data in natural language, while more than 300 pre-trained extraction models handle common document types without training.
Customers choose their models: Tungsten describes Bring Your Own Model support that combines leading providers such as Azure OpenAI with a customer's own custom LLMs. Agents reach external tools through MCP, and MCP governance decides which tools each agent may use while logging that use in full. Tungsten's own robotic process automation drives application screens, identifying labels, buttons and text boxes, including inside Citrix, and AI answer quality can be optimized over time against standardized benchmarks.
TotalAgility Cloud holds FedRAMP High authority to operate, announced in March 2026, and meets U.S. state and local and Australian government security requirements, alongside SOC 2 Type II. The platform also runs on-premises and in hybrid models, so agencies moving off solely on-premises deployments can adopt the cloud alongside their existing systems.
Vendor details
Canonical URL
https://www.tungstenautomation.com
Category
Enterprise operations agent
Subcategory
Agentic automation on an intelligent document processing platform
Funding status
Private, headquartered in Irvine, California. Formerly Kofax, renamed Tungsten Automation in 2024.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
TotalAgility agents invoke external tools and AI models through MCP, with MCP governance controlling which tools each agent may use and logging that use in full. Tungsten's own robotic process automation drives application screens, including in Citrix environments, and knowledge bases load documents of any size for retrieval. A DocAI Studio SDK supports building document AI solutions.
In practice
A bank processes garnishee orders end to end, with agents classifying the document, extracting the obligation and driving the downstream case without an operator opening each file.
A federal agency inherits security controls from the FedRAMP High boundary to shorten its own authorization cycle, moving document workflows off purely on premise deployments into cloud or hybrid.
A compliance team queries years of unstructured case files in natural language through the Question Answering Agent rather than commissioning a manual document review.
Sources & related URLs
Research sources
Agentic Index coverage score
13.0 / 14 capabilities · 93%
| Integrations & Tool Calling | Full |
|---|---|
|
Agents reach outside systems through a documented interface. Expanded MCP integration, new in TotalAgility 2026.3, lets agents invoke external tools through standardized interfaces, with the agent consuming MCP as a client. Sourcetungstenautomation.com/products/totalagility/release-highlightsread 2026-09-18 |
|
| Workflow Orchestration | Full |
|
Multi-agent orchestration is an available TotalAgility capability that coordinates multiple AI agents across enterprise workflows, alongside orchestration of document-driven workflows across enterprise systems. Tungsten's knowledge discovery example shows worker agents collaborating with each other in natural language and structured JSON inside a workflow that combines traditional automation with agent steps. Sourcetungstenautomation.com/products/totalagility/release-highlightsread 2026-09-18 |
|
| Knowledge Grounding & RAG | Full |
|
Agents answer from knowledge bases the customer loads. Any document, regardless of size, can be loaded into a knowledge base in a single automation step, with chunking, OCR and vectorization preparing it, and the AI agent uses the knowledge base in a retrieval-augmented generation pattern to locate relevant information. AI Knowledge Bases centralize and search unstructured enterprise content, and natural-language queries return traceable, source-backed insights. New material enters by loading it, not by retraining. Sourcetungstenautomation.com/blog/building-an-intelligent-knowledge-discovery-agentread 2026-09-18 |
|
| Human Oversight & Guardrails | Full |
|
Tungsten names human-in-the-loop controls and exception handling as built-in TotalAgility capabilities, with SLA management, so exceptions route to a person inside the workflow rather than being resolved by the agent alone. Agents run inside case-driven process boundaries. Approvals sit at the workflow layer. Sourcetungstenautomation.com/blog/building-inhouse-vs-partnering-with-an-automation-expertread 2026-09-18 |
|
| Security, Identity & Governance | Full |
|
TotalAgility Cloud carries government authorizations listed on Tungsten's release page, meeting U.S. federal security requirements, U.S. state, local, tribal and educational government requirements, and Australian federal, state and territory government security requirements. Tungsten's trust center states adherence to SOC 2 Type II, NIST, HIPAA and GDPR. The customer-facing control on the agent layer is enhanced MCP governance, which lets the customer control which AI tools each agent may access, and AI interactions are clearly identified for transparency. Sourcetungstenautomation.com/products/totalagility/release-highlightsread 2026-09-18 |
|
| Observability & Auditability | Full |
|
What agents do is recorded. TotalAgility logs agents' use of AI tools in full under its MCP governance, and Tungsten names audit trails and continuous monitoring among the platform's built-in capabilities, so each agent's tool access and the workflow it runs inside leave an auditable record. The behavior on the record is the agent's. Sourcetungstenautomation.com/products/totalagility/release-highlightsread 2026-09-18 |
|
| Memory & State Persistence | Partial |
|
Users query case data through natural-language interactions with TotalAgility's agents, which carries conversation state within a session. The case record the agent reads is the business record itself, the application's data model rather than a memory the agent keeps, and knowledge bases are reference material the customer loads. No agent memory layer with its own scope and lifetime is documented, so nothing says whether memories can be reviewed, edited, deleted or scoped by user, team or workspace. Sourcetungstenautomation.com/products/totalagility/release-highlightsread 2026-09-18 |
|
| Deployment & Data Residency | Full |
|
TotalAgility runs on the customer's own premises as well as in TotalAgility Cloud. Tungsten's FedRAMP High announcement of 19 March 2026 describes agencies moving from solely on-premises deployments to secure cloud and hybrid models, using TotalAgility alongside their existing applications. Federal agencies can reach TotalAgility Cloud directly or through public-sector contract vehicles. Sourcetungstenautomation.com/press-releases/2026/fedramp-high-tungsten-totalagility-cloud-shortens-agencies-path-to-secure-process-modernizationread 2026-09-18 |
|
| Prebuilt Agents, Templates & Packs | Full |
|
Tungsten ships customer-adoptable packaged assets. More than 300 pre-trained extraction models across industries and document types deploy with no training required, and Tungsten cites more than 3,000 pre-trained models across its ecosystem. Each model does its own job on its own document type, so removing one leaves the rest whole. Sourcetungstenautomation.com/products/totalagility/release-highlightsread 2026-09-18 |
|
| Triggers & Channel Coverage | Full |
|
Work reaches TotalAgility without a person asking. The platform is described as a foundation for document intake, classification, routing and task automation, so an arriving document is classified and routed into the process on its own rather than on a user's request. Document intake is the event-driven trigger. Sourcetungstenautomation.com/press-releases/2026/fedramp-high-tungsten-totalagility-cloud-shortens-agencies-path-to-secure-process-modernizationread 2026-09-18 |
|
| Model Flexibility & Routing | Full |
|
The customer chooses the model. Tungsten states that TotalAgility's Bring Your Own Model capabilities let customers combine leading AI providers, including Microsoft Azure OpenAI Service and others, as well as their own custom LLMs. That is customer control of the model rather than vendor-side routing. Sourcetungstenautomation.com/blog/building-inhouse-vs-partnering-with-an-automation-expertread 2026-09-18 |
|
| APIs, SDKs & MCP Extensibility | Partial |
|
Builders extend TotalAgility with a DocAI Studio SDK for deploying document AI solutions and with AI-assisted development for creating agents, both listed on the release page. The platform's MCP support is consumed by its agents as a client, which makes it an integration rather than a way to call TotalAgility from outside. No public API reference for invoking TotalAgility agents or processes from outside is documented. Sourcetungstenautomation.com/products/totalagility/release-highlightsread 2026-09-18 |
|
| Testing, Debugging & Optimization | Full |
|
Answer quality can be optimized over time against standardized benchmarking, one of TotalAgility's knowledge discovery capabilities listed as available on the release page. A standardized benchmark gives a comparable score to optimize against, though the benchmark sets and result format are not documented. Sourcetungstenautomation.com/products/totalagility/release-highlightsread 2026-09-18 |
|
| Browser & Computer Use | Full |
|
Tungsten ships its own robotic process automation that operates real application interfaces. Its robots capture screen images and locate and identify UI elements such as labels, buttons and text boxes within application environments including Citrix, communicate with virtualized applications, and run concurrently under centralized robot management with version control across development, test and production. That is documented control of a real interface, not a scrape or a fetch. The robots cope with screen elements by identifying them on the rendered screen. Sourcetungstenautomation.com/products/rparead 2026-09-18 |
|
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
Recent platform changes
Tungsten Automation launched InvoiceAgility+, an invoice to pay platform where dedicated AI agents check invoices for fraud patterns and duplicates before payment, predict GL codes and cost centers for invoices without a purchase order, and translate invoices in more than 100 languages. Every agent recommendation carries a confidence indicator and a short explanation, thresholds are adjustable, and items that need judgment go to human review.
Bears on: Human approval / guardrails
View sourcePricing
Contact sales; three named tiers published without prices
tiered platform packages aligned to use case
Included quota
Not published. Tier entitlements are described qualitatively rather than as quotas.
What is public
The three tier structure and its use case alignment are published. No prices, quotas or agent consumption units are published.
Billing mechanics
Not publicly disclosed. Sold direct and through a partner channel including Titanium tier partners, and through US public sector contract vehicles for federal buyers.
Cost watchouts
Tier boundaries are drawn around capability rather than volume, so a customer needing knowledge discovery over unstructured content is pushed to the top tier regardless of size. Document processing volume and the number of RPA bots are the historical cost drivers in this category. TOKEN COST IS THE LIVE RISK: a Tungsten customer publicly calculated that a fully agentic approach to one AML and KYC workload would run 5 million USD a year in tokens against 250,000 for a hybrid design, so agent architecture choices dominate run cost far more than license tier.
Variable cost rationale
Nothing commercial is published, so cost cannot be forecast from vendor pages. More importantly this is the one vendor in the sweep with hard evidence of how severe agent run cost can be: a customer calculated 5 million USD a year in tokens for a fully agentic AML and KYC design against 250,000 for a hybrid alternative. The variable component is therefore driven by architecture decisions made after purchase, which is the least predictable kind of exposure.
Additional watchouts
Model the token cost of the agent design separately from the license. Tungsten's own customer evidence shows a twenty fold swing between fully agentic and hybrid architectures on the same workload.
Overage / add-ons
Not published.
Sales call required
Yes, required for paid access
Free / trial
Demo on request
Lowest paid plan
Intelligent Document Processing (Standard)
Commercial notes
Private company with roughly 25,000 customers and over 350 federal agencies. Scale is well evidenced even though pricing is not.
Key ambiguities
All prices, plus how agent execution is metered. Tungsten publishes no named agent consumption unit, so whether agent runs draw on a license entitlement or on pass through model cost is unknown.
Missing data
List prices, tier entitlements as quotas, agent metering, and RPA bot licensing.
Related vendors
- 11th Estate — Agentic platform whose AI engine scans markets, matches an…
- Adopt AI — AI CPA firm whose agents run reconciliations, close, AP and AR, and…
- 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…
- Alloy.ai — Commerce intelligence system for consumer brands that unifies four…
Alternatives to Tungsten Automation
The closest documented capability profiles to Tungsten Automation 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.
- Pega13.5 / 14Fuller documented coverage on APIs, SDKs & MCP Extensibility
- Appian14.0 / 14Fuller documented coverage on Memory & State Persistence and APIs, SDKs & MCP Extensibility
- Atomicwork12.0 / 14A lighter documented profile than Tungsten Automation
- Leena AI12.0 / 14A lighter documented profile than Tungsten Automation
- Microsoft14.0 / 14Fuller documented coverage on Memory & State Persistence and APIs, SDKs & MCP Extensibility
- ServiceNow14.0 / 14Fuller documented coverage on Memory & State Persistence and APIs, SDKs & MCP Extensibility
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded