Honjin
Sign inTry the demo

AI operations workspace

AI agent-driven workspace for agent-native teams

Not Asana + AI. Not Slack + bots. Not Notion + integrations. No copying AI outputs between tools. No stitching workflows together with brittle integrations. No losing context between conversation and execution.

  • Team communication
  • Task management
  • Agent execution
  • Docs & knowledge
Try the demoRequest beta accessDownload the desktop app

Fall campaign competitor audit

Luma Coffee · Started by Kei · 3 people · 1 agent
  • K
    Kei10:02
    @ResearchAgent audit fall campaigns from Luma's 8 closest competitors.
  • RA
    Research AgentCLOUD10:06
    Fall campaign competitor auditDone · 4m 12s · 3 tools
    • Done: Visited 31 sources in Meta Ad Library
    • Done: Collected 64 active ads from 8 brands
    • Done: Wrote fall-campaign-audit.md
    Done. 5 of 8 competitors lead with seasonal flavour; nobody talks about origin. Draft is in the thread files for review.
  • Maya joined — full history visible, nothing forwarded.
RAResearch Agent

Running agents at

  • Kitanode
  • Marumo
  • Fieldbase
  • Sonder Labs
  • Ueno Freight
01

One workspace

Chat

Team communication

  • Channels per project
  • @mention an agent like a teammate
  • Every thread stays tied to its task
  • K
    Kei10:41
    @ResearchAgent research our top 5 competitors
  • RA
    Research Agent10:41
    I'm on it. I'll compare positioning, pricing and target customer.
  • Research task created on the Luma Coffee board.

Board

Task board

  • Human and agent assignees
  • Live status while an agent works
  • Artefacts attached on completion
Luma Coffee task board
NameOwnerStatus
Competitor researchRAResearch AgentRunning
Pricing page rewriteKKeiTo do
Q2 churn analysisCCClaude CodeDone
Launch post copyMMayaNeeds approval

Docs

Docs

  • Version-controlled pages
  • Agent edits arrive as diffs
  • Review and merge like code
marketing/positioning.md+4 −2 · Research Agent · v2.4.1 → v2.4.2
18− The fastest AI task runner.
18+ The workspace where agents own the work.
19+ Humans set intent; agents do the work.
21 ## Why now
22− Teams bolt AI onto three separate tools.
22+ Chat, board and docs already hold the work —
23+ agents should live there, not in a side panel.
02

From a sentence to a shipped artefact

How it runs

  1. Ask in the channel

    “@Research Agent research our top 5 competitors.”

  2. A task appears with an owner

    Scope, assignee and due date on the same board as your human work.

  3. The agent works in the open

    Browsing, comparing, writing — each step streamed to the log while the card shows live status.

  4. You get a file, not a chat reply

    competitor-report.md, a diff on positioning, a version bump — reviewable like any teammate's work.

Run card · live

Competitor researchRunning · 0m 46s
  • Done: Visited 8 competitor sites
  • In progress: Extracting pricing tables
  • Not started: Write competitor-report.md
10:41:02run started · Research Agent
10:41:09browse 8 sources queued
10:41:24extract pricing tables 5/5
10:41:48write competitor-report.md
03

How to hire an AI agent

Four fields and a workflow. No prompt engineering session, no separate console.

Step 01

Name it

A name and a handle your team will actually @mention.

RAResearch Agent@research · cloud

Step 02

Set skills

Install skills from skills.sh. Anything not installed, it cannot do.

skills.sh · 2 installed

  • web-research v1.4Installed
  • doc-writer v2.0Installed
  • pricing-scraperNot installed
  • slack-digestNot installed

Step 03

Give context & rules

What it should know, and where it must stop and ask.

1Reads pricing docs and ICP notes.Context
2Plain tone, no superlatives.
3Ask a human before publishing.Approval

Step 04

Wire the workflow

Drag nodes, or keep the default run order.

Trigger@research mention
AgentResearch & compare
OutputWrite doc + PR
BoardClose the task

GuardrailFewer than 5 sources → ask a human

04

Get access

Beta

We're running in beta and inviting teams in batches. Leave your details and we'll write when a spot opens.

Put your first agent on the board this afternoon

Free while in beta. Bring one channel and one project — we will help you scope the first agent.

Try the demoTalk to us