One assistant that knows your company's documents.

Ask in Slack, Teams or a widget on your site and get an answer from your own policies, contracts, manuals and tickets, with the source attached. People only get answers from documents they may open. It runs on the models and hosting you choose, including fully on your own servers.

answer comes with its source and a link
Every
to the first assistant over your documents
2–4 wk
of your access rules respected, from SSO to folders
100%

Company assistant

Answering in Slack, with sources

ask-hr
D

Question from Deniz

What is our notice period for contractors in Germany?

Searching 1,284 documents · permissions checked

Sources

  • 1HR policy Germany 2025.pdf · page 4
  • 2Contractor agreement template.docx · clause 9
  • 1 document skipped: no access for Deniz

Answer

  • Contractors in Germany have a 30-day notice period after the first three months [1].
  • During the first three months either side can end the contract with 7 days' notice [2].
  • Notice must be in writing; email to the contractor and HR counts [1].

Was this right? Yes No

Answered in 2.1 s

HR, support and IT leads put an AI knowledge base on their documents.

Same assistant, same process. Which source and which channel go first depends on who is asking.

Head of HR / Operations

Before
Half my day is answering the same policy questions in Slack. The answer is in a PDF nobody opens.
After
The assistant answers in the channel where people ask, from the current policy, with the page attached. You see which questions keep coming back and fix the document instead of the inbox.

2–4 wkto the first assistant live

Support lead

Before
Agents search four tools before they can answer a customer. New hires take months to learn where things are.
After
One search over the help center, the internal wiki, past tickets and the product docs, with the source attached. Agents paste a checked answer instead of hunting for one; new hires are useful in their first week.

Everyanswer with its source and a link

Head of IT / Legal

Before
We tried a chatbot and it answered from documents people were not allowed to see.
After
Access is checked at question time against your SSO and the source system's own permissions. Hidden documents stay hidden, admins see which documents were used in every answer, and the index runs on hosting you choose.

100%of your access rules respected

You own the AI knowledge base we build.

  • 01

    Connectors to your sources

    Google Drive, SharePoint, Confluence, Notion, Zendesk, Jira, HubSpot, file shares and databases. PDFs, Office files, wiki pages, tickets and emails indexed and kept in sync as they change.

  • 02

    Permission-aware answers with citations

    Every answer is built from retrieved passages and names the document and page. Access is checked at question time, and when the documents do not have the answer, the assistant says so.

  • 03

    Slack, Teams, widget and API

    The assistant answers where people already ask: a Slack channel, a Teams chat, a widget on your intranet or site, and an API your own tools call.

  • 04

    The admin dashboard and the evals

    Unanswered and low-rated questions, sources used per answer, usage per team and an eval suite on 100 of your real questions that runs before every change.

Included free

4 documents ship with the entry package.

The same folder you see on the pricing page. Open any one to read a sample first page.

All of these documents come free with the package; none of them adds a line to your plan.

  • Need something else?Just ask. We add it to your package, free.

Your AI knowledge base answers in two to four weeks.

You bring the documents, 100 real questions and an hour a week. We connect, index, measure and launch, and your team checks the answers before anyone else sees them.

  1. 01

    Week 1

    First meeting and source map

    We
    A free first meeting about who asks what, and where the answers live. Then a research report: sources, access model, hosting, pilot questions and a fixed price.
    You
    Approve the plan and pay 50%. Collect 100 real questions from your team.
  2. 02

    Weeks 2–3

    Connect and index

    We
    Connect the first source, build the index on your hosting, map permissions to your SSO and build retrieval with citations. Write the evals from your questions.
    You
    Review the first answers with us and mark what is wrong.
  3. 03

    Weeks 3–4

    Launch in one channel

    We
    Slack, Teams, the widget or the API goes live for a pilot group. Tune retrieval on their feedback, then open it to everyone.
    You
    Pick the pilot group and announce it in the channel.
  4. 04

    After launch

    Add sources and improve

    We
    Review unanswered questions weekly, add the next source and channel, re-run the evals. The remaining 50% is due at delivery.
    You
    Fix the documents the dashboard flags, or send them to us.

Start with your documents. Grow into company-wide search.

Three packages, each at a fixed price you see before you commit. Each one builds on the one before, so nothing is thrown away.

  1. 01

    First meeting

    30 minutes, free

    You tell us which questions keep coming back and where the answers live. We tell you which source and channel to start with and what the entry package costs.

  2. 02

    Document assistant

    Live in 1 to 2 weeks

    Your team asks; the assistant answers from your documents, with the source.

  3. 03

    Company knowledge base

    Live in 2 to 4 weeks

    Every source in one place, and each person sees only what they may.

  4. 04

    Enterprise search

    Live in 4 to 8 weeks

    A knowledge base for thousands of people, with single sign-on and an audit trail.

See the package

The AI knowledge base reads your documents and answers in your chat tools.

Connectors to the document, ticket and chat tools you already run, on the model and hosting you choose. Anything with an API gets a connector we hand over.

Anything else with an API: we write the connector and hand you the code.

Documents and wikis

  • Notion
  • Gmail
  • Outlook
  • Google Sheets
  • Airtable
  • Jira
  • GitHub

Tickets and CRM

  • Zendesk
  • Intercom
  • Freshdesk
  • HubSpot
  • Salesforce

Where it answers

  • Slack
  • Microsoft Teams
  • WhatsApp
  • Telegram

Models and hosting

  • OpenAI
  • Anthropic
  • PostgreSQL
  • Supabase
  • AWS
  • Azure

We built these products that answer from documents.

Legally

A legal document app with an AI chat that answers questions about a scanned contract, clause by clause, across 13 languages and 16 jurisdictions. Serving thousands of users who have scanned hundreds of thousands of documents.

500,000+documents scanned

techsy.io
Browser content

Akollo

Our product, live demo

Our enterprise work platform with an assistant that answers from projects, budgets and documents, reads only what the signed-in user may see, and logs every AI action in a chained audit trail. Installed in the customer's own data center.

Use it with any AI tool.

The agents, skills and code we hand over work in the tools your team already uses.

  • Codex
  • Claude
  • Cursor
  • Grok
  • Z.ai
  • Hermes
  • OpenClaw
  • Eve
  • Mistral
  • OpenCode
  • VS Code
  • Antigravity
  • Windsurf
  • JetBrains
  • Zed
  • GitHub Copilot
  • Cline
  • Qwen
  • Amp
  • Goose
  • Lovable
  • Replit
  • v0

Your AI knowledge base package has one fixed price.

  • Lowest price

    Document assistant

    Your team asks; the assistant answers from your documents, with the source.

    $1,200

    $800

    Starting price · one-off payment

    • 4 documents, free
    • Up to 500 documents from one source, such as Drive, SharePoint or Notion
    • Every answer shows the page it came from
    • Team chat for up to 25 people
    • Updates itself when a document changes
    • A setup guide and a short training session
  • Most popular

    Mid-range

    Company knowledge base

    Every source in one place, and each person sees only what they may.

    $4,500

    $3,000

    Starting price · one-off payment

    Everything in Document assistant, plus:

    • All your sources connected: drives, wikis, tickets and email
    • Permissions follow the source
    • Inside Slack or Teams, plus a version for customers
    • Up to 250 users
    • Quality tests on real questions

  • Top of the range

    Enterprise search

    A knowledge base for thousands of people, with single sign-on and an audit trail.

    $13,000

    $9,000

    Starting price · one-off payment

    Everything in Company knowledge base, plus:

    • Thousands of users across companies and languages
    • Single sign-on and an audit log of every answer
    • An API so your other systems can ask it too
    • Quality tests and monitoring on every change
    • Hosting sized for your load, in your region

Your first meeting with Techsy is free.

You leave understanding what you need.

Book the free first meeting
  • The first meeting costs nothing.
  • The research report is free too.
  • 25% back if you cancel within 28 days.
How a project runs with Techsy, step by step
  1. Free meeting: Thirty minutes with a person who builds.
  2. Planning: Scope, phases and dates, agreed in writing.
  3. Research and docs: A free research report and the specs.
  4. Solution and benefits: What we build and what it saves you.
  5. Development: Built in short cycles you can see.
  6. Launch: Live, monitored and handed over.

Teams ask these questions before putting an assistant on their documents.

Where the data goes, who sees what, how accurate it is and what it costs. The questions from every first meeting, answered here.

  • Three packages. Document assistant for $800 connects one source for up to 25 people and shows the page behind every answer. Company knowledge base for $3,000 connects all your sources with their permissions, inside Slack or Teams. Enterprise search for $9,000 serves thousands of users with single sign-on and an audit log. You pay 50% when you approve the plan and the rest at delivery. Model and hosting costs run in your own accounts at the provider's price.

  • Where you decide. The index and the documents stay in your cloud account or on your own servers; we never copy them to ours. The model can be an API with a no-training agreement and a data processing agreement, or an open model on your hardware so nothing leaves your network at all. The research report names every component and where it runs, so your data protection officer can sign off before we build.

  • No. Access is checked at question time against your identity provider and the source system's own permissions. If a document is hidden from someone in SharePoint, Google Drive or Confluence, it is hidden from their answer too, and the assistant says that it found nothing it may show. Admins see which documents were used in each answer, so the rule is auditable.

  • Every answer is built from retrieved passages and cites them, so you can check it in one click. When the documents do not contain the answer, the assistant says so instead of guessing. Before launch we run evals on 100 real questions from your team and agree on a threshold; nothing goes live below it. The dashboard lists unanswered and low-rated questions after launch, which is usually a sign of a missing or outdated document.

  • Google Drive, SharePoint and OneDrive, Confluence, Notion, Jira, Zendesk, Intercom, HubSpot, Salesforce, GitHub, file shares and databases through their APIs. Anything else with an API gets a connector we write and hand over. PDFs, Word and Excel files, slides, wiki pages, tickets and emails are all indexed, including scanned PDFs after OCR.

  • Yes. The assistant answers in a Slack channel or direct message, in a Teams chat, in a widget on your intranet or website, and through an API your own tools call. Most teams start with one channel and add the others after launch. The answer format follows the channel: short with links in chat, longer with a source list in the widget.

  • Yes. People ask in their language and get an answer in it, even when the document is in another one. A German policy answers a Turkish question, with the source attached. We test the languages your team actually uses in the evals, so the threshold covers them.

  • The first meeting is free. Within a week you have a research report with the sources, the access model, the pilot questions and a fixed price. You approve the plan and pay 50%. One source and one channel are live in two to four weeks; all sources, all channels and SSO take up to three months. The remaining 50% is due at delivery, after the evals pass the threshold we agreed.

Didn’t find your answer?

Ask a person

Ship the system your business runs on.

Book a scoping call and we map the build, the cost, and the timeline. Most teams leave the first one with a fixed plan.

Book a 30-min scoping call

Three quick questions, then your price band and timeline. No call needed.