PromptQL Tag | The company-wide AI agent for SlackSign in

Sign inBook demoTry PromptQLpath]:stroke-[#030712] ml-1" width="23" height="22" viewBox="0 0 23 22" fill="none" xmlns="http://www.w3.org/2000/svg">

PromptQL Tag<br>One AI agent for your whole company. Tag it in Slack.

Not a custom agent per channel, but a company-wide AI that respects security & privacy. Tag it in any channel and it works alongside your team — acting with each person's own permissions, running on any LLM, and building a shared brain as it goes.

And it's your AI — your brand, your models, your context, your identity and on your cloud.

Get started→Request demo

Play 90s demo

Trusted at scale by

Trusted at scale by

It's yours<br>It's not Claude's. It's yours.<br>Your brand, your models, your context, your identity and on your cloud.

Brand<br>Not Claude's name & face.<br>It's YOUR brand.

span]:!h-7 [&>span]:!w-7 [&>span]:!rounded-[9px] [&>span]:!text-[9px]">ClaudeApp<br>span]:!h-7 [&>span]:!w-7 [&>span]:!rounded-[9px] [&>span]:!text-[9px]">NWNorthwindApp

Your company AI has your identity. Especially when you have guests.

Models<br>Not locked to Claude's models.<br>It's YOUR models.

Claude<br>GPT<br>DeepSeek<br>Gemini<br>GLGLM

Choose and switch between a variety of models, especially for images, video, audio.

Context<br>Not Claude's black-box memory.<br>It's YOUR context.

Claude memory

Your wiki

An internal "Wikipedia" for your team. With version history and notifications.

Security<br>Not Claude's agent identity.<br>It's your users' identities.

SSarafull<br>FFinancescoped<br>AAcme guestexternal

personal: SaraFinanceAllcustomer: Acme

Each user's real permissions, enforced at the data layer.

Cloud<br>Not Claude's cloud.<br>It's YOUR cloud.

Your cloudsingle-tenant<br>VPCBYOCOn-prem

Runs in your own cloud — single-tenant VPC, BYOC, or on-prem. Your data stays put.

Every teammate keeps their own permissions. No new rules to set up.<br>The agent acts as whoever tagged it, inheriting that person's access to data and tools — so there are no channel-wide credentials and no dumbed-down lowest-common-denominator bot.

Acts with the triggering user's permissions; authorization follows the real person.<br>Row-level and column-level access enforced deterministically at the data layer — the model can't bypass it.<br>Audit logs attribute every action to the actual human, not a shared bot identity.<br>Learn more about user identity→

revenueMSNWthe agent acts as whoever tagged it

MMaya9:14 AM<br>@Northwind what's our Q1 revenue by region?

NWNorthwindApp9:14 AM<br>Running as Maya (Finance) — full access:<br>Q1 revenueFinance<br>AMER$26.0M<br>EMEA$18.4M<br>APAC$9.2M

SSara9:15 AM<br>@Northwind same question for me — what's our Q1 revenue by region?

NWNorthwindApp9:15 AM<br>Running as Sara (APAC sales) — scoped to your access:<br>Q1 revenuepersonal: Sara<br>AMER•••<br>EMEA•••<br>APAC$9.2M

AMER & EMEA hidden by row-level access before the model sees them.

Access scopespersonal: SaraFinanceAllcustomer: Acme

Any model. Even for images.<br>Pick the best model for each task — frontier, open-weight, or multimodal — without ever leaving Slack.

Any LLM, per channel or per task — no lock-in.<br>Route cheap or open-weight models to control cost.<br>Ask it to generate an image (e.g. “use Nano Banana”) right in the thread.<br>Any model, any modality<br>Claude<br>GPT<br>DeepSeek<br>Grok<br>Nano Bananamedia<br>GLGLM

launch-campaignMANWany model, per channel or per task<br>Nano Banana

MMaya7:12 PM<br>Use Nano Banana to make a launch image — dark background, neon-green mascot, room for a headline.

NWNorthwindApp7:13 PM<br>Switched this thread to Nano Banana and dropped the render inline below.

AAri7:14 PM<br>Love it. Now switch back to a cheap text model and tighten the launch copy.

NWNorthwindApp7:14 PM<br>Back on an open-weight text model for edits — same thread, no context lost.GLMcheaper for copy edits

It builds your company's shared brain as you work.<br>Correct it once and the fix sticks — for everyone, in every future thread. A shared, validated wiki, not private per-channel memory.

Captures decisions and corrections as human-reviewed, cited wiki entries.<br>One shared, versioned, cross-linked brain — knowledge doesn't fragment per channel.<br>The next teammate gets the right answer without re-explaining it.<br>Learn how shared context works→

customer-successMDNWcorrections become shared, cited wiki entries

MMaya11:20 AM<br>Acme's support tickets just spiked. Are they a churn risk?

NWNorthwindApp11:20 AM<br>Usage is steady and NPS is healthy, so on the numbers alone Acme looks low risk.

DDana11:22 AM<br>Not quite — their champion left in March and renewal is in May. For Acme, low usage is high risk, not low.

NWNorthwindApp11:22 AM<br>Got it — capturing that as shared context so the whole team starts from it.<br>Team AI wants to learn<br>Review once, save once — every future thread starts from it.

Your wiki · your cloud<br>Account: Acme<br>Champion left in March; renewal is in May.<br>Treat low usage here as high churn risk.<br>Loop in the assigned CSM on any support spike.

Add to wikiEdit<br>Enterprise: hosted in your...