Ask. Build. Manage.
Employees ask in Chat. Builders create agents and workflows in Studio. IT runs everything from one console. All three share the same knowledge, the same integrations and the same rules.
Chat · Every employee · Ask questions, get answers with sources, and let agents do tasks.
One shared foundation: Knowledge · Integrations · Trust layer. Ready-made business solutions that install into all three at once.
Ask · Chat
An assistant that shows its work.
Employees pick the right agent and ask. Answers stream in, grounded in your documents, with the sources one click away.
- Every answer names the manual, section and page.
- You see each search the assistant ran.
Cited answers
Every answer lists its sources with excerpts and page numbers. Open one to see the actual page.
The right agent for the job
Favourites and recent agents, each with its own welcome and example questions, in Finnish and English.
Pin documents as context
Pin documents that are already indexed to a conversation. Access is checked again on every message.
Agents that do things
Search knowledge, the web and databases, work with Jira and GitHub, send email. You see each step.
Approve in the conversation
Sensitive actions wait for your Approve or Reject, right in the chat.
Stop, archive, find again
Stop an answer mid-way. History is grouped by time, searchable, and can be archived and restored.
Ask · Memory
Remembers what matters. With permission.
Assistants can remember your role, preferences and ongoing work, and recall earlier conversations. Opt-in, visible and editable.
- Each fact shows where it came from.
- Opt-in: switch memory or auto-learning off.
You decide
Memory is opt-in. Turn it on, turn off automatic learning, or mark a conversation as Don't remember this.
See every fact
Each remembered fact shows where it came from: added by you, saved by an agent or learned automatically.
Private by design
Encrypted at rest, personal data masked in summaries, every change audited. Recall only searches your own history.
Ask · Knowledge
Search that knows who is asking.
huoku reads documents and data from your systems and makes them searchable, in your languages, with the original access rules intact.
- Sources grouped into one knowledge base.
- Which Spaces and agents may use it.
Keyword and meaning together
Finds exact terms and ticket keys as well as answers phrased differently. Optional AI re-ranking for relevance.
Permissions in the search itself
Results are filtered by the asker's rights inside the search engine, not only in the app.
Multilingual, Finnish included
The language model that indexes your documents runs inside the platform. Nothing is sent out to be indexed.
Real-world documents
PDFs, Office files, spreadsheets, tables and scanned pages with text recognition.
Knowledge bases
Group sources by topic and decide which agents and teams may use each one.
Related material, optionally
Can recognise people, projects, systems and products across documents to find what belongs together.
Build · Studio
Build agents without writing code.
Studio is where builders in each team create agents, skills and workflows, with an assistant that drafts them from a plain description.
- Starts on its own: a schedule, a webhook, a Kafka event, a new document or a Slack message.
- Pipeline, mesh or supervisor: choose how agents work together.
Agent builder
Instructions, model, tools, knowledge and which actions need approval. Test it before anyone else sees it.
Skills
Reusable know-how, like your outreach voice or how to triage an alert, that agents pull in when needed.
Visual workflows
Chain agents on a canvas: fixed steps, a team that hands work around, or a lead agent directing specialists.
An assistant for builders
Describe what you want. It fills in the form and you confirm before anything applies.
See what happened
Watch runs live, stop them, and follow each step. Sensitive content stays hidden from builders.
Documentation as you build
Every agent gets a versioned system card, written with the EU AI Act in mind.
Build · Automation
Workflows that start themselves, and wait for you.
Workflows run on a schedule or when something happens. Anything sensitive pauses for a person, then continues exactly where it stopped.
- The run pauses until a person decides.
- Resolve the approval and it continues where it stopped.
Starts on its own
Schedules, webhooks, polling, Kafka, new documents, Slack mentions, GitHub events, incoming email.
Does real work
Sends email, creates Jira and GitHub issues, queries databases, calls APIs, posts back to Slack.
Pauses safely
Approval pauses the whole workflow, even across several agents, and resumes it cleanly.
One approvals queue
Background approvals collect in one place with a pending count. Decisions can be remembered per workflow.
Alerts when work waits
Email when approvals go stale or pile up.
Bounded by design
Step and time limits on every workflow. Anything running can be stopped.
Foundation · Packages
Business processes, ready to run.
A package brings the agents, workflows and work screens for one process. Activate it, answer a short setup, and it runs.
- A package brings its own work screen.
- Versioned, licensed and switched on per package.
Guided setup
A short setup form. An assistant can pre-fill it from your website.
Updates you review
Every update is shown as a change list before it applies, and can be rolled back.
Safe by default
Nothing that talks to the outside world is on until you configure it. Disabling never deletes.
Foundation · Integrations
Connects to what you already run.
One page to connect your systems. Each connection can bring knowledge, tools for agents, and events that start workflows.
Work
- Jira
- Confluence
- Slack
- Microsoft Teams
- GitHub
Files and web
- SharePoint
- Windows file shares
- Local directories
- Web crawler
Business data
- PostgreSQL
- MySQL
- SQL Server
- Oracle
- Snowflake
- SAP (OData)
Events
- Inbound webhooks
- HTTP polling
- Kafka
- Incoming email
Infrastructure
- Kubernetes
Platform services
- Email sending
- Web search, with a European option
Knowledge, tools and triggers
One connection can feed search, give agents actions and start workflows.
Test before you trust
Tested from where it will run, even inside your network. Wrong password, missing permission or unreachable: it tells you which.
Credentials stay put
Encrypted, and cleared if the address changes, so a secret never follows a moved endpoint.
Teams come along
Groups and roles imported from Jira, GitHub, Teams and Confluence.
Manage · Spaces & access
One platform, many teams.
Spaces give Sales, HR, Legal or Maintenance their own area: their own builders, agents, workflows and chats, invisible to the others.
- Builders for this Space only.
- Service accounts for automation.
Builders per team
Grant build rights inside one Space without platform-wide access.
Hidden means hidden
What belongs to another Space doesn't show up at all, search results included.
Automation without borrowed identities
Workflows run as named service accounts that can never hold admin or builder rights.
Single sign-on
Your identity provider, with Microsoft group sync and LDAP.
Manage · Platform management
One console for IT.
Packages, integrations, models, access, approvals, monitoring and audit, in one place. Your brand, your models.
Choose your models
Per agent: European-hosted or open-weight models via any OpenAI-compatible endpoint, your own via vLLM or Ollama, or the major providers.
Monitor quality and cost
Usage, agent activity, user feedback, infrastructure health, and per-model speed and cost.
Your brand
Product name, colours, logos and favicon, changed at runtime.
Infrastructure as code
A Terraform provider manages agents, workflows, models and identity.
Personal-data rules
Data Guard patterns are tested live, then roll out to every instance at once.
Audit built in
Every access change, configuration change and agent action on record.