One brain. For you, and your whole company.
Neuron captures how your company actually works and serves verified answers to every person, and every AI agent, you run.
neuron · ask anything
Redis Cluster was chosen Mar 2023 after the Feb 2023 rate-limiter incident, Memcached's LRU evicted live token buckets under burst load past 1M keys.
Trade-off accepted +$1,800/mo infra cost · zero token-bucket loss
Decision owner @priya (former Staff Eng, left Aug 2024)
⚑ Conflict detected, frozen
infra/README.md line 88 still says "use Memcached for ephemeral counters." Last updated 2022. Owner pinged.
Sources · 3
- doc
Original decision doc
RFC-0042-rate-limiter.md · Mar 2023
- incident
Feb 2023 1M-key eviction PIR
post-incident-117 · Feb 2023
- slack
Priya's tradeoff thread
#infra-eng · Mar 11, 2023
Served to
Same verified row · served to your team in Slack, your agents over MCP, and your morning brief
Plugs into the tools you already live in
Slack
Notion
GitHub
Gmail
Drive
Claude
OpenAI
Cursor
HubSpot
JiraSlackNotionGitHubGmailDriveClaudeOpenAICursorHubSpotJira<1s
Verified answer
across Slack, web, and MCP
3x
Faster onboarding
from months to days
0
Stale answers served
frozen on conflict, never silently wrong
10+
Tools, one brain
Slack, Notion, GitHub, Gmail, Linear, MCP
How it's used
One brain. Four ways to use it.
01 / 04
Just me
Your second brain that doesn't make you organize anything. Captures from your browser, email, notes, and calendar.
- Browser, email, notes, calendar captured passively
- Morning brief written while you sleep
- No team setup needed
Personal brain · Ali
liveDecisions yesterday: 3 Blockers overnight: 2 Things you forgot: 5 > what did I tell @samira at Linear last week?
02 / 04
My team
Onboarding goes from months to days. Every Slack rule, Notion doc, and decision is one query away.
- Slack, Notion, GitHub, Linear synced continuously
- Verified answers under a second
- New hires productive in days
Team brain · 24 members
live> why is staging 4x slower than prod? From postmortem (May 2025): staging shares a Postgres replica with analytics. @rafa proposed splitting; not done.
03 / 04
My company
When sources contradict, Neuron freezes the rule, pings the owner, and only serves verified answers.
- Conflict detection across every source
- Owner pinged automatically
- Audit log of every change
Company brain · conflict resolved
live> what's our refund window? 14 days. Verified by @marie, Jun 03. Frozen 1 conflict: support macro from Q4 still says 30.
04 / 04
My agents
Your agents query the same brain your team does, over MCP. Claude, ChatGPT, and Cursor never hallucinate on company context.
- Three MCP tools: query, context, save
- Same verified row everywhere
- Workspace-scoped, never trains a public model
MCP · query_company_brain
live{
"answer": "Postgres is the source of truth;
Redis is read-cache only.",
"confidence": 0.94,
"sources": ["adr-0009", "slack/#infra"]
}Why Neuron is different
Other tools search.
Neuron decides what's true.
When two sources contradict, Neuron freezes the rule, pings the owner, and only serves verified answers. Closer to Git than to Google.
Two sources, disagreeing
3-year term ending 31 Jan 2026, with 90-day notice required to avoid auto-renew. Negotiated $0.04/credit floor in exchange for case-study rights.
@sarah · head of data
Snowflake: 1-year term, auto-renew, no notice required.
@intern-rosa
Snowflake renewal, 3-year term.
Auto-renew unless you give 90 days' notice. Deadline: 2 Nov 2025.
Pinged · @ops, @cfo
no stale answer servedAsk anything your team already knows
One brain. Every question answered.
Under the hood
Listens silently. Extracts structure.
Serves verified answers.
Listens
Connects to Slack, Notion, GitHub, Gmail, Linear, your browser. Continuous, never trains on your data.
Extracts
Pulls rules, decisions, processes, ideas, each with source, owner, and date.
Serves
Web. Slack /neuron. MCP server in Claude Desktop. Same verified row, three surfaces.
Proactive, not just reactive
Your morning brief, in the inbox you already check.
Each morning Neuron tells you what shifted overnight and which decisions are at risk. No new app to learn.
InboxNeuron brief@neuron.so
3 decisions yesterday. 2 at risk. 1 blocker.
Built from 47 messages, 12 docs, and 6 calendar events overnight.
What moved
- Switched to Stripe Tax after Marie's Q1 memo.
- Q3 mobile scope cut by 2 features. Linear MOB-218.
- Vendor decision: Adyen retained for EU. Rationale logged.
Needs you
- Snowflake renewal sign-off, deadline 2 Nov.
- Refund window: Marie set 14d, support macro still says 30. Pinging @ops.
3 things you said you'd do
- Intro to @samira at Linear (promised Tue).
- Review the refund policy doc.
- Board prep deck — empty since Friday.
What gets brewed overnight
- 47Slack messages parsed
- 12Docs cross-referenced
- 6Calendar events linked to decisions
- 3Conflicts detected and frozen
Roles tuned for
- Founders — what shifted, what you committed
- Eng leads — what's blocked, where the conflicts are
- Sales leads — pricing changes, customer commitments
- Ops / People — vendor renewals, policy edits
Delivery
Email at 06:00 local. Slack DM at 09:00. Both source the same verified row your team queries during the day.
Integrations
One brain. Every surface.
Every tool your team uses flows into one verified brain.
Neuron
MCP-native from day one
The first company brain your agents
can actually trust.
Glean is bolting MCP on. We were built around it. Three tools. Same row your team queries in Slack.
Claude Desktop · neuron-mcp
(question: string)
-> { answer, confidence, sources[], verified }Ask anything. Returns the verified answer plus the underlying sources and a confidence score. Never returns a frozen item.
3 tools. Same row as /neuron in Slack.
Live MCP call
> query_company_brain("what is our refund policy?")
streaming...
0/10
Passive capture, on your terms
Silently captures, never silently shares.
Zero training.
Your data never trains a public model. Ever.
Channel-level controls.
Choose what Neuron reads. Exclude private channels, DMs, or specific repos.
Workspace-scoped.
Knowledge is isolated per workspace. Engineering can't see HR; HR can't see legal.
Retention you own.
Set forget-after-N-days per source. Delete anything anytime.
The bet
Companies are getting amnesia at the speed of AI.
Every Slack thread, every Cursor session, every Linear ticket is a decision that will be wrong in three months and gone in twelve. Search tools surface them. We adjudicate them.
The next decade will run on AI agents querying company context. Those agents will be wrong as often as the docs they read. Neuron is the verified context layer underneath them.
Abdi & Ali, founders
Currently in beta
08/20
design partner slots filled
We're onboarding teams that run on Slack and ship with agents.
Partners get hands-on engineering, custom MCP tools, and a direct line to the founders. 12 slots left this quarter.