The first generation of AI was personal. The next will be collaborative.
One reality. Shared by people and AI.
Your team and all its agents work from the same pages: handbooks, skills, decisions. Every contribution builds on the last, and you stay in control.
- For you, your team and your organization
- Works next to the tools you already use, like Claude Code, Codex and Copilot
- Hosted in Switzerland, made for Europe
Sound familiar?
Keeping a team on the same page was already hard enough.Now everyone brings their own agents.
- Your knowledge
- Your team reads the handbook. Your agents read a copy someone wrote for them months ago, and every new chat starts from zero.
- Sharing
- Your SKILL.md and DESIGN.md work well in your setup. Sharing them with colleagues and their agent systems means copying files around, and the copies drift apart.How skills stay in one place ↓
- Unchecked instructions
- Your agents follow instructions nobody has checked in months.
- Different agents
- Anna works with Claude Code, Marco with Codex. Each finds something different, and neither knows what the other found.
How it works
Three steps to a shared start.
Start with what you already have
Write it, import it, or let your agents gather it.
Bring your people and collaborate
Write, review and decide together. Write down the why, and every agent knows it too.
Connect your agents and collaborate
Claude Code, Codex, Copilot or your own. They work in the same pages as your team, and build on each other’s work.
Your team’s understanding lives here: notes, skills, decisions, analyses. Keep your files where they are, and link to them.
Who it is for
For you, your team and your organization.
For you
Your notes, with your agents
Notes, sources and ideas, linked as you go. Claude and Hermes work in them for you, and remember what you worked on yesterday. Nobody else sees them.
Control
Keep control. See what your agents do.
More of the work is done by agents. You see what they changed, what they use, and where they need help.
Discount rules for 2027
Skills
Write a skill once. Every agent uses it.
A skill tells agents how your team does something: how a quote is written, how a report is checked. Written in the open SKILL.md format, kept in one place.
- You can’t share what lives on one laptop.
- One place Each skill lives once in your library, not as copies in every tool. Claude Code, Codex and Copilot read the same version.
- You can’t trust what nobody approved.
- Approved A person switches each version on. A skill an agent writes arrives switched off.
- You can’t improve what you can’t measure.
- Flagged when stale People and agents rate every skill. You see when one stops helping.
Good work leaves a skill behind: an agent writes it down, a person switches it on.
quote-templateSKILL.md
- Written by Lea’s agent switched off
- Approved by Lea switched on
- Used by Claude Code, Codex and Copilot rated 4.5
- Rating dropped 2.7Prices moved to the 2027 list.Reported by 3 agents · summarized by HontoFlow
- Version 2: new prices waiting for approval
Built for agents
Your agents find the right page.In any language.
Your pages are in German, French and English. An agent asks in Italian. Honto hybrid search still finds the right page, fast and precise.
- Agents connect through the Model Context Protocol (MCP), the open standard Claude Code, Codex, Copilot and Cursor already use.
- Words, meaning and links in one ranking.
- In our benchmark, the right page was in the top three 95 to 97 % of the time, in each of seven languages.
- Everything your agents find carries its version, its rights and its feedback.
For organizations
Built for how organizations work.
Access down to a single page
Viewer, Editor or Admin, per library, folder or page. No agent reaches more than its person.
A history with authors
Each change shows who made it. Go back in the history when you need to.
Hosted in Switzerland, made for Europe
Your data is stored in Switzerland. The built-in agents use AI models in Switzerland and Europe. Connect the agents you trust.
No lock-in
Your knowledge, your rules and what your agents learned belong to you. Export it all in open formats, any time: Markdown, Open Knowledge Format, SKILL.md.
Traces for agents, privacy for people
See what your agents use: every read leaves a trace. What your people read is never tracked.
Routines and loops, within budget
HontoFlow’s own agents run routines and loops for you, on a schedule. Every loop has a goal, limits and a budget you set.
Our story
Made for looping.
We at Hontolab built HontoFlow to close our own loops. A loop is one question that agents work on for weeks, run after run. Different agents take turns, and we steer between the runs. That only works when everyone, people and agents, is on the same page.
So we kept using it. Today our whole team and all of our agents work in it together, across all our processes, from a blog post to the technology we build.
Every run starts from what the last one learned. We call it Continuous Inquiry.
