Ask anything. Answered
from your team's truth.

Lumen reads every doc, ticket and Slack thread your company has ever written — and gives one trustworthy answer, with sources, in under two seconds.

Explore your spaces Browse the library⌘K to ask

184,920

Documents indexed

412k

Answers served

1.3s

Avg response

Connectors

Live answer

1.2s · 6 sources

"How do I rotate a production API token?"

Generate the replacement token in the Vault dashboard (Settings → Tokens → Rotate). The old token stays valid for 24h to avoid downtime. Update the value in 1Password under `prod/api-keys`, then trigger the `deploy:tokens` workflow — it patches all running services in <90s.

Grounded in

Postgres connection pooling — incident retroEngineering · Doc
RFC-0184 · Event bus deprecationEngineering · GitHub
Onboarding 2.0 — checklist redesignPeople Ops · Doc

↑ Helpful ↓ Not quite Lumen v3.2 · gpt-class router

Connectors

42 sources

NotionGoogle DriveSlackGitHubLinearConfluenceJiraFigmaZendeskIntercom+32 more

Permission aware

Answers respect the same ACLs as your source tools.

SAML · SCIM · per-document scope

Top spaces this week

View all →

⚙
Engineering
1,284 docs · 3,120 answers
2m ago
◇
Product
612 docs · 1,487 answers
9m ago
✦
Go-to-Market
488 docs · 902 answers
14m ago

P95 latency, last 24h

1.3s−0.4s vs last week

Streaming-first router · 14 regions

Recently indexed

Open library →

EngineeringDoc
Postgres connection pooling — incident retro 11/14
PgBouncer saturated session pool during the marketing launch; here's the new transaction-mode rollout plan.
12 minutes ago
Go-to-MarketNotion
Q1 pricing experiment — early signals
Self-serve conversion lifts +14% on the 3-tier card. Annual toggle has surprisingly low adoption.
1 hour ago
People OpsDoc
Onboarding 2.0 — checklist redesign
New hires complete day-one setup in 38 min on average (was 2h 10m). Equipment kit changes attached.
3 hours ago

How it works

From scattered docs to a single answer in four moves.

No data leaves your tenant. We index, embed, and retrieve — your model of choice drafts the final answer.

Connect

Pick from 40+ first-party connectors. OAuth in one click, scoped to read.

Index

Lumen chunks, embeds, and links everything — preserving ACLs end to end.

Ask

Type in plain English. We route to the right space and synthesise an answer.

Trust

Every claim is cited inline; one tap takes you back to the source paragraph.

Features

Built for teams that read more than they write.

Permission-aware retrieval

Lumen mirrors the access controls of every source it indexes. A Support agent and a CFO ask the same question and get answers grounded in different documents — automatically.

Source-first citations

Every sentence is anchored to a passage. Hover for a snippet, click to jump to the doc in its original tool.

Smart routing

A question about 'rotation' hits Engineering. About 'parental leave', People Ops. Lumen learns this from your taxonomy in under a week.

Workflow shortcuts

Slash-commands and a global ⌘K. Ask Lumen from anywhere — Slack thread, Linear comment, browser address bar — and pipe the answer right back.

In their words

Teams quietly love it. Then they tell us.

"Onboarding used to take three weeks of pestering teammates. Lumen cut it to four days. Hires answer their own questions before standup."

Hana Okafor

Head of People, Northwind

"Our on-call rotation finally stopped paging the same two engineers for the same five tribal-knowledge gaps. Lumen quietly absorbed the load."

Priya Natarajan

Staff Engineer, Helios

"We retired three internal wiki search projects after a single quarter. The ROI is unambiguous and the answers actually cite where they came from."

Marcus Lee

VP Product, Foundry

Changelog

Shipped, quietly, every week.

A small selection from the last month. The full feed lives in-product.

v3.2
Jun 24
Streaming citations
Sources resolve as the answer is being drafted — no more 1.4s 'thinking' gap before sources appear.
v3.1
Jun 17
Bring-your-own model
Route inference through your AWS Bedrock or Azure OpenAI deployment with one config switch.
v3.0
Jun 03
Permission-aware Spaces
Spaces now inherit ACLs from your IdP. Same question, different answer, same security model.
v2.9
May 27
Linear + Jira
Two more connectors. Ticket history is now first-class context for engineering questions.

Security

Built to pass the procurement review.

SOC 2 Type II, ISO 27001 in audit, region pinning across US/EU/AU, single-tenant VPC on Org. Your data stays where you put it.

SOC 2 Type II
ISO 27001 (in audit)
SAML SSO + SCIM
Region pinning
Bring your own KMS
PII redaction filters
Audit log streaming
No training on your data

FAQ

The questions you'd ask, anyway.

How long does indexing take?

A 10,000-doc workspace finishes a cold index in roughly 18 minutes. Incremental syncs run continuously and propagate edits in under a minute.

Will Lumen train models on our content?

No. Your documents, prompts and answers are never used for training, fine-tuning, or evaluation. This is enforced contractually and at the infrastructure layer.

Can we self-host?

Yes — Lumen ships as a single-tenant VPC deployment in AWS, GCP or Azure, with region pinning. Available on the Org plan.

What if a permission changes?

ACL deltas are picked up at the same cadence as your IdP — typically within seconds. Out-of-date answers are invalidated and re-grounded before being served.

What models does Lumen use?

A router selects between frontier and small models per question. You can lock to a single provider or route through your own deployment on Org.

What does it integrate with?

42 first-party connectors today, with a public webhooks API for the long tail. Most teams are productive within the first hour.

System status

All systems normal

99.98% uptime · 90 days

14 regions · 1 brief degradation on May 04 (12 min, EU-west)

Careers

We're hiring 7 roles

Distributed-systems engineers, an applied researcher, and a designer who reads books. Remote in EU/US.

Privacy first, last, always

Your knowledge is yours. Lumen is the lens — never the library.

Delete your workspace and every embedding, vector, and cached document is purged within 30 days. We'll send you the receipt.

Bring the answers.
Leave the searching behind.

Get started See live insights

Ask anything. Answeredfrom your team's truth.