Model Agnostic AI Coworker

Instub is an AI coworker that lives where your team collaborates. It has full context on your company, so it takes actions and real work off your team's plate.

Real conversations in Slack

Slack

Instub

AGENT

Direct message

InstubAGENT9:04 AM

Hi Maya — two things I'd set up this week: a Monday recap in #sales, and a Wednesday invoice chase. Start both?

M
Maya9:05 AM

yes both

InstubAGENT9:05 AM

@Maya Both are on. First recap Monday at 9, in #sales.

ConnectorsHubSpotGmail

Doing real work at

Context.dev28 ROWConiferopflowThunder ComputeAudun

Four parts, one coworker

GitHubStripeZendeskNotionPostHogYour DB

Cloud agents that live in your team chat

Instub runs on its own machine in the cloud and works where your team operates. Every run is sandboxed, scoped to the tools you connected, and torn down when it finishes.

Awaiting your call

Run this every Monday at 9?

Yes, schedule itNot yet

Automations

Nobody has to sit down and configure workflows. When Instub sees the same request three Mondays running, it offers to take it over.

pricing.mddeal-desk.mdrefund-policy.mdonboarding.md

A wiki that organizes itself

Not a folder of documents that rot. The pages write themselves out of how the company actually operates.

Built for a whole team, not one user

Everyone in the channel works with the same Instub, but nobody borrows anyone else's access. Permissions follow the person who asked.

# growth
PostHog
Stripe
# Stripe

Your next hire has agency

Instub doesn't wait to be told. It volunteers for the jobs it sees repeating, posts finished work with the steps behind it, and holds anything that leaves your company until you sign off.

It asks for the work

The third time someone pulls the Monday numbers, Instub offers to own the job: same run, every Monday at 9:00, posted to #growth. You approve once.

MON 04Audit CAC by campaign
MON 11Audit CAC by campaign
MON 18Audit CAC by campaign
Awaiting your call

Run this every Monday at 9?

✓ Yes, take itNot yet

How the trust model works →

Your knowledge stops walking out the door

Every decision your company makes is already written down, scattered across channels nobody searches. Instub reads it, remembers it, and keeps the wiki current so the answer exists before anyone asks.

Wiki · refund-policy.md · revised

Updated 2 min ago · reason: policy change in #ops

− Refunds over $500 need approval from the deal desk.

+ Refunds over $1,000 need approval from the deal desk. Amounts under $1,000 are auto-processed within 24 hours.

#ops · today 9:42

revised by Instub · 3 people notified · previous version archived

One tag. Eighty models.

Most tagged work does not need a frontier model. Instub routes it through Conifer to whichever of more than eighty fits, and reaches for Claude, GPT, or Gemini only when the task needs them.

  • Claude Fable 5
  • Claude Opus 5
  • Claude Opus 4.8
  • Claude Sonnet 5
  • Claude Sonnet 4.6
  • Claude Haiku 4.5
  • GPT-5.6 Sol
  • GPT-5.6 Terra
  • GPT-5.6 Luna
  • GPT-5.5
  • GPT-5.4 mini
  • gpt-oss 120B
  • Gemini 3.1 Pro
  • Gemini 3.6 Flash
  • Gemini 3.5 Flash
  • Gemini 3.5 Flash-Lite
  • Gemini 3.1 Flash-Lite
  • Grok 4.5
  • Grok 4.3
  • DeepSeek V4 Pro
  • DeepSeek V4 Flash
  • DeepSeek V3.2
  • Kimi K3
  • Kimi K2.7 Code
  • Kimi K2.6
  • Qwen3.7 Max
  • Qwen3 Coder 480B
  • Llama 4 Maverick
  • Llama 4 Scout
  • Mistral Medium
  • Mistral Small
  • GLM-5.2
  • GLM-4.7
  • MiniMax M3
  • Nemotron 3 Ultra
  • Nemotron 3 Super
  • Command A
  • Gemma 4
  • Seed 2.0 Pro
  • Seed 2.0 Mini
  • Hunyuan 3
  • Step 3.7 Flash

80+

models to route to

70%

lower model spend

1

mention to reach them

FAQ

The tagging layer shouldn't belong to a model company.

Model-agnostic. Faster. Aligned with you.

Add Instub to Slack or Microsoft Teams and tag a task. It routes to the model that should do the work and comes back in-thread.

Rather be shown around first? Book a 15-minute intro.