Kora
Start Free

Hand work to AI. Without losing control.

Systems run the fixed steps, AI agents judge within bounds, people decide the exceptions, every run recorded.

Start FreeBook a Demo

Interactive Kora console preview

Chat

What should Kora work on?

When an invoice arrives, match it to the purchase order, clear the exact matches, and raise anything over 2 percent variance to finance.
§ 01 · Workflow

Improve your processes with run data you own.

Define a process in plain language and release it as a governed version. Watch every run as it happens. The replayable record you own is audit-ready evidence and shows what to change in the next version.

Governed workflow canvas: services, AI agents, people, and connectors assembled into a released workflow, with a loop feeding run observations back into the next version.A canvas labeled Your Organization. Icons for databases, APIs, documents, model providers, roles, and messaging tools sit in four connected cards, Services, AI Agents, People, and Connectors, inside a workflow frame stamped with a version and released today, under a green governed badge listing observability, audit logs, and proof. A dashed loop routes the workflow output back to its start, resolving into a lamp badge on a card headed suggested improvements: auto-approve the routine path, retry the flaky connection, and drop the unused step, feeding a next candidate version. On load the icons first appear scattered across the canvas before organizing into this workflow.YOUR ORGANIZATIONGOVERNED · OBSERVABILITY · AUDIT LOGS · PROOFWORKFLOW · V1.4RELEASED · 2026-09-14SUGGESTED IMPROVEMENTSAuto-approve the routine pathRetry the flaky connectionDrop the unused stepNEXT: V1.5 · CANDIDATEServicesDatabaseAPIMCPDocumentsAI AgentsAnthropicGPTMistralGLMPeopleAnalystQAReviewerSecurityConnectorsGmailOutlookSlackTeamsGoverned workflow canvas: services, AI agents, people, and connectors assembled into a released workflow, with a loop feeding run observations back into the next version.YOUR ORGANIZATIONGOVERNEDWORKFLOW · V1.4RELEASED · 2026-09-14SUGGESTED IMPROVEMENTSAuto-approve the routine pathRetry the flaky connectionDrop the unused stepNEXT: V1.5 · CANDIDATEServicesDatabaseAPIMCPDocumentsAI AgentsAnthropicGPTMistralGLMPeopleAnalystQAReviewerSecurityConnectorsGmailOutlookSlackTeams
§ 02 · Governance

Governed, observed, audit ready.

Hand critical work to AI and automation, without handing over control, or the proof. You always know the true state of every run, prove it the moment anyone asks, and fix a broken one in minutes, not days.

WORKFLOW · V1.4 · RUN 4022PROD
OBSERVABILITY · TRACES
01

Operational certainty

Every run is visible while it happens: which step is live, which actor holds it, what was just decided. The state of the operation is a fact on screen, not a guess assembled from check-ins.

AUDIT LOG · HASH-CHAINED
02

Evidence for auditors

When a customer, operator, or internal review asks how a decision was made, the recorded evidence already answers: what ran, who approved it, under which released version. Evidence is a lookup, not a scramble.

REPLAY · EVENTS
03

Debugging in minutes

A failed run replays step by step with the exact state it ran under. Root cause stops being detective work, and the fix can be proven against the very run that broke.

§ 03 · Authoring

Bring what you have. The agent builds the workflow.

Process maps, SOPs, spreadsheets, and the exceptions nobody wrote down. Attach what exists, describe the rest in plain language, and the agent writes both halves: the versioned definition and the code that talks to your systems. You review it before release; nobody wires up nodes by hand.

KoraClaudeCodex

What should Kora build?

process-map.bpmn24 KBapprovals-sop.pdf180 KBdata-schema.xlsx41 KB

Turn this BPMN and our SOP docs into a governed workflow.

This shift is not ours alone.

You won't have to use different apps for different tasks.

Bill GatesBill GatesCo-founder, MicrosoftGatesNotes · 2023

The IT department of every company is going to be the HR department of AI agents.

Jensen HuangJensen HuangFounder and CEO, NVIDIACES keynote · 2025

Build for agents.

Andrej KarpathyAndrej KarpathyFounding member, OpenAIAI Startup School · 2025

The most universal interface: natural language.

Satya NadellaSatya NadellaChairman and CEO, MicrosoftAnnual letter · 2023

You could write programs in English, and the computer would translate them into machine code.

Grace HopperGrace HopperComputing pioneer, US NavyOn the first compiler

Build software entirely using natural language.

Amjad MasadAmjad MasadCo-founder and CEO, ReplitOn Replit Agent · 2024
§ 04 · Connectors

Connected to the systems already in place.

Workflows act through the tools the organization already runs. Every connector, prebuilt or built for you, operates under the same governance as any participant: scoped permissions and recorded evidence on every call.

CONNECTORSHundreds of prebuilt integrations
Salesforce
SAP
Oracle Database
SharePoint
Outlook
Jira
Confluence
BigQuery
Teams
OneDrive
Zendesk
GitHub
Google Sheets
Slack
Gmail
100+ morePrebuilt integrations
Custom connectorsLet Kora build one designed for your own stack, systems, and needs.
§ 05 · Deployment

Versioned, tested, released, reversible.

Every version is a release: an immutable package of the definition, its steps, and its policy. Deploying applies it to a named environment, and rolling back means redeploying the release that ran before.

Deployment pipeline

  1. Definequeuedrunning240ms
  2. Testqueuedrunning12.4s
  3. Releasequeuedrunning180ms
  4. Deployqueuedrunning3.1s

Stages

4 steps · all green

  1. 01
    Define

    Describe the work in chat or code. Kora compiles it into a versioned workflow definition your team can review.

  2. 02
    Test

    Kora validates the definition before it can be released, so a broken process never becomes a package.

  3. 03
    Release

    Cut a release. Inputs, steps, and policy freeze into an immutable, versioned package.

  4. 04
    Deploy

    Apply the release to a named environment. A failed attempt leaves the live one untouched, and rollback redeploys an earlier release.

§ 06 · Enterprise

Keep Kora in your environment.

Enterprise runs Kora in an environment your organization selects and controls. Your team operates the workflows; Kora can manage or support the deployment when included in the agreed Enterprise services.

Talk to Sales
Isometric enterprise infrastructure with four redundant modules surrounding a customer-controlled workflow core.

Your control

Your organization selects the environment, keeps workflow data there, and operates the workflows.

Kora support

When agreed, Kora can manage or support the deployment and help prepare it for production.

Self-host where you need it.

§ 07 · Solutions

Where teams apply governed workflows.

The same platform for repeatable work across teams. Start from a use case or the outcome you are after.

Use cases
A buyer's hands reviewing supplier paperwork with a pen at a desk, pausing on one order before approving it.
01

Procurement exceptions

Automate routine replenishment, surface price, quantity, and supply exceptions, and keep a buyer in control before any purchase order is placed.

A maintenance technician holding a rugged tablet beside an industrial pump skid, one amber status light glowing on the equipment panel.
02

Maintenance escalation

Kora turns maintenance and production alerts into governed, AI-assisted workflows: it gathers context, ranks urgency, routes supervisor approval, and dispatches the approved work order to your CMMS or EAM with a traceable decision record.

An accounts payable specialist holding a paper invoice at a bright desk, a tray of invoices and a softly blurred finance screen beside it.
03

Invoice processing

Run accounts payable as a governed workflow where an agent matches and validates each invoice, and a person approves the exceptions before anything is paid.

An operations specialist reviewing a queue of request sheets beside two work screens and a stack of case folders.
04

Request triage

Gather the details, classify the request, and route it through one released workflow, with people deciding the cases that need judgment.

A customer operations lead reviewing an onboarding packet at a desk before approving the next handoff.
05

Customer onboarding

Turn the kickoff, setup tasks, dependencies, and customer updates into one workflow with a clear owner at every exception.

An order operations specialist reviewing an exception list beside a live operations screen.
06

Order exception handling

Bring order, inventory, fulfillment, and customer context together before an operations owner chooses the resolution.

By outcome
An operations desk running unattended: a monitor showing a workflow of connected steps, most cleared with a green check and one banded in amber, a printed process sheet on a clipboard beside the keyboard.
01

Governed AI Workflows

Connect people, systems, and AI agents into governed workflows that produce reliable, auditable business operations.

A hand marking a revision in pencil on a printed standard work procedure pinned to an operations board, a run chart with a rising trend line pinned beside it, a plant floor softly blurred behind.
02

Operational Excellence

Kora turns manual procedures into governed workflows your organization owns: AI agents run the routine path, people decide the exceptions, and every run is measured so the process keeps improving.

§ 08 · FAQ

Questions, answered.

Straight answers to what teams evaluating governed workflows ask: what Kora is, how it differs, where it runs, how data is handled, what it connects to, and how it is priced.

Kora is a workflow operating system for organizations that run important business processes across people, systems, approvals, and AI. People, AI agents, and systems execute each process together under a released, governed version, and every run leaves replayable evidence: Platform Audit records control-plane actions, Runtime Evidence explains execution, and the authoritative event feed exports ordered, hashed workflow facts for audit.

Systems run the fixed steps, AI agents judge within the bounds the released workflow sets, and anything outside those bounds is raised to a person with the rule result and supporting evidence attached. Every run, step, and decision lands in an ordered event record with a digest per fact, so any run can be opened and shown. The interactive console on this page walks through one process: describe it in chat, connect the mailbox and the ERP it uses, release and deploy it, watch five runs arrive, decide the one raised to finance, and read its record.

Workflow governance means an organization controls how its business processes execute: who can change a process, which version runs, who approves exceptions, and what evidence each run leaves behind. It extends workflow automation with ownership, policy, and auditability.

An agentic workflow is a business process where AI agents, people, and systems execute the steps together toward an outcome, with each agent able to take actions, call tools, and make decisions inside the process. In Kora, agentic workflows run under a released, governed version, and every run leaves a detailed, replayable evidence record.

Kora treats the workflow itself as a governed asset, while automation tools focus on executing tasks. Your organization owns the process definition, controls releases, observes every run, and expands automation over time only where the run evidence supports it.

A process is described in conversation, and Kora compiles and validates that description into a versioned workflow definition, then presents the workflow graph as a read-only view to inspect. Teams can also author from an assistant they already work in, such as Claude or Codex, and every change still reaches production through release control.

Yes. AI agents work alongside people and systems in a workflow. Each agent runs on a model chosen from Kora's supported list, across providers you likely already use, including Anthropic's Claude and OpenAI's GPT models, plus Google, xAI, Groq, Mistral, DeepSeek, and more, with an operator supplying the provider API key per environment. Custom and self-hosted model endpoints are on the roadmap. Each agent runs under the released process version, acts only within the permissions granted to its step, and leaves a recorded, hash-chained trace for every action.

Kora SaaS is hosted and operated by Kora. Enterprise is the self-hosted option for an organization that needs Kora in an environment it selects and controls. The exact deployment model is scoped to its infrastructure, security, data residency, identity, integration, monitoring, and support requirements.

In Kora SaaS, workflow data is stored and processed in the Kora-hosted environment serving your account. With self-hosted Enterprise, workflows and evidence stay in the environment you choose and control. The Terms and Privacy Notice describe the data handling that applies to each service.

Yes. Self-hosted Enterprise supports data sovereignty by keeping Kora in an environment your organization selects and controls. Kora or your team can manage the deployment, while your team builds and operates workflows inside Kora.

Kora connects to hundreds of prebuilt integrations across the enterprise stack, including Salesforce, SharePoint, Outlook, Microsoft Teams, Jira, Confluence, Box, Zendesk, GitHub, Google BigQuery, and Google Sheets. It also reaches internal APIs and databases through service operations that run governed code, extracts content from documents in a workflow step, connects to Streamable HTTP MCP endpoints, and can build a custom connector for your own stack. Every connector runs under the same scoped permissions as any other step, and its calls are recorded in the run evidence.

Every production run is captured as a durable, hash-verifiable record, tagged with its release, environment, and deployment, so the data is attributable to the version that produced it. Every change already ships through release control: immutable releases, controlled environment deployment, and audited rollback by redeploying an earlier release. Replaying a candidate version against those historical runs, and gating promotion on how it compares, is on the active roadmap and not yet enforced today.

That is the direction. Kora keeps cross-referenced evidence layers: a control-plane audit log of who changed what, plus runtime evidence, agent traces, and a journal of the workflow's business facts. The same captured run data an audit relies on is what an evaluation loop would later consume, so no extra instrumentation is needed. Using those records to backtest candidate versions, and to export governed datasets to a fine-tuning or reinforcement-learning provider you choose, is on the roadmap, not live today. Kora does not train models itself, and where that data lives follows your deployment.

Yes. On Kora SaaS, the first user is free and each additional user is $49 per month. AI credits are purchased separately on demand. Enterprise is a custom license for organizations that want Kora on their own infrastructure. Full details are on the pricing page.
§ 09 · From the blog

Latest from the Kora team.

Workflow governance in practice: product thinking, platform engineering, and lessons from real processes.

Read all posts on the Kora blog
§ 10 · Next step

One real process is enough.

In a demo, we take a process you run today, describe it in chat, run it under governance, and show the record it leaves: evidence for auditors, and the data that makes the next version better.

Book a DemoStart Free