Kora
Book a DemoStart Free

Agentic workflows with audit-grade reliability.

Kora turns critical business processes into executable workflows where humans, systems, and agents operate under release control, runtime observability, and evidence capture from every run.

Start FreeBook a Demo
§ 01 · Workflows

Work Kora can run for you.

Routine checks move automatically. People review exceptions. The proof is already there.

01Factory work

Product checks

Routine batches clear themselves. People see only the unusual cases.

GainMistakes caught earlyOut-of-spec batches surface before release.

02Quality work

Issue follow-up

Findings get owners, actions, approvals, and a clean history.

GainIssues resolved fasterOwners, actions, and approvals stay connected.

03Operations

Supplier and order checks

Files, missing approvals, and mismatches are sorted before they slow the line.

GainManual checks reducedClean orders move without waiting.

04Finance and governance

Morning review packs

Overnight breaks and model changes are prepared before teams start the day.

GainMorning prep savedTeams start with exceptions already sorted.
§ 02 · Works with any AI

Describe the work. The workflow builds itself.

Plug Kora into the AI your team already uses. Define the outcome in chat; Kora plans the steps, connects the right tools, and prepares the workflow for review before release.

ChatGPTCodexClaudeKora
kora

Send me a request from any connected chat. I'll turn it into a workflow.

you

Right now when a supplier emails their inspection.csv, we manually check each batch against our tolerance table, then ping the quality lead in Slack and email for approval. can you turn this into a workflow?

kora
Planning workflow
Gmail triggerinspection.csv
Python steptolerance check
Slack + email approval@quality-lead
kora

Workflow ready. Inspect each node, then publish a release.

supplier-deviation.workflowready
Triggergmail

Email trigger

inspection.csv arrives from supplier

Coderuntime

Python · tolerance check

load inspection.csvjoin tolerance_tableemit deviation_report
Human stop@quality-lead

Quality lead approval

slackemail
process ledgertriggercode outputhuman approval
01 / CLI

Available from the CLI.

Kora can be operated from the command line, so agents and developers can create, inspect, and publish workflows where automation work already happens.

One surface. Terminal, chat, and workflow state stay aligned.
02 / Skills

Models know how Kora works.

Kora ships skills that teach AI models the commands, guardrails, and release path. The model can operate the system directly instead of asking users to memorize screens.

Embedded know-how. Tools · policies · workflow actions.
03 / Adoption

No product training loop.

Users describe the outcome in plain language. The model uses Kora skills to plan the work, choose the right operations, and bring back a workflow that is ready for review.

Less learning. Explain the work, not the software.
§ 03 · Reliability

Open any run. See what happened. Prove it again.

Every step, every input, every output is recorded and hash-chained the moment it happens. Open a run to inspect it. Replay it to prove it ran the same way. Hand the file to your auditor without rebuilding history.

run.WF-1148

supplier-deviation · production

Hash-chaineddigest 9c2f88b1f4a2…
Observed6 steps · 1m 53.142s
  1. gmail120ms
  2. python340ms
  3. human approval1m 50.8s
  4. branch8ms
  5. sap245ms
  6. commit117ms
Logged6 hashed events
  1. 14:02:41gmail.received8b3f…
  2. 14:02:41python.tolerance.checke22a…
  3. 14:02:43human.approval.request—
  4. 14:04:33human.approval.granted9f3a…
  5. 14:04:34sap.procurement.update—
  6. 14:04:34evidence.commit—
Auditableappend-only · hash-chained

approved by alice@kora

  1. requestedslack + email
  2. grantedalice@kora
  3. recordedevidence record

policy · quality.gate.v3

Replayablescrub · play · prove again
  1. gmail120ms
  2. python340ms
  3. human approval1m 50.8s
  4. branch8ms
  5. sap245ms
  6. commit117ms
00:01:42 / 01:53.142

Replay verifiedsame digest as the recorded run

§ 04 · Deployment

Workflows ship like software.

Define the spec. Cut a release. Test the gates. Deploy to production. Kora treats every workflow like a release artifact, versioned and replayable, so the path from idea to production looks like the one your engineers already trust.

Deployment pipeline

Define · Release · Test · Deploy

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

Stages

4 steps · all green

  1. 01
    Define

    Describe the work in chat or code. Kora produces a versioned workflow spec your team can review.

  2. 02
    Release

    Cut a release. Inputs, steps, and policy frozen into an immutable, versioned envelope.

  3. 03
    Test

    Gate the release: dry-runs, replay against history, policy checks, before anything reaches production.

  4. 04
    Deploy

    Roll out to your runtime: on-premises, EU-sovereign cloud, or hybrid. Every run is hash-chained.

Plugs into your CI/CDGitHub ActionsGitLab CIJenkinsCircleCIArgo CDBitbucket Pipelines

§ 05 · Where Kora runs

Runs in your world, not ours.

Kora runs on your own infrastructure, on-prem or air-gapped, as a self-contained deployment you bring to whatever you already operate. Kora Cloud manages accounts, licensing, and instances, and stays out of your workflow's critical path. Plans add quotas and connectors on top.

View PricingTalk to Sales

The runtime is always free. Quotas and connectors scale by tier.

§ 06 · FAQ

Questions, answered.

Straight answers to what teams ask about Kora: what it is, how it works, where it runs, and what it costs.

Kora is a platform for building and running agentic workflows with audit-grade reliability. It turns critical business processes into executable workflows where people, AI agents, and code work together under release control, runtime observability, and a hash-chained, replayable evidence record on every run.

You describe the outcome in plain language from the AI assistant your team already uses. Kora plans the steps, connects the right tools, and produces a versioned workflow spec to review. From there you cut a release, test it against policy and replay gates, and deploy to production, so workflows ship like software.

Each agent runs on a model from Kora's supported list, including Anthropic's Claude and OpenAI's GPT, plus Google, xAI, Groq, Mistral, and DeepSeek, with an operator supplying the provider API key per environment. You can pin or swap the model for each workflow. Custom and self-hosted model endpoints are on the roadmap.

Every run is observed, logged, and hash-chained into an append-only evidence record the moment it happens: step timings, hashed inputs and outputs, and the human approval chain. You can open any run, replay it to the same cryptographic digest, and hand the record to an auditor without rebuilding history from logs. Signing and anchoring are handled by a downstream system.

Kora is a self-contained deployment that installs from a Docker Compose bundle and runs wherever you operate: your own data centre, a cloud VM in AWS, Azure, or Google Cloud, or fully air-gapped. A Kubernetes target is on the roadmap. There is no SaaS tenant and no Kora-hosted control plane in your critical path.

Your workflows run inside your environment, and an operator supplies the provider API key per environment. Cloud-connected non-Enterprise plans send Kora Cloud telemetry and data-contribution records described in the Terms and Privacy Notice; Enterprise and offline deployments do not participate in Kora Cloud telemetry or data contribution.

Kora has a free tier plus Pro, Teams (priced per seat), and Enterprise plans. Every plan runs the same runtime; paid plans raise the licensed limits, seats, organizations, and live workflows, and Enterprise adds offline licensing, custom limits, and compliance addenda. See the pricing page for the full comparison.

Kora produces the audit-grade records regulated teams need, including for DORA (binding since January 2025) and the EU AI Act Annex III evidence model (from 2 August 2026). Enterprise agreements add a DORA Chapter 5 addendum covering audit rights, exit strategy, concentration-risk disclosure, and incident reporting.