Live Context
The knowledge layer

Company knowledge that's
actionable, not just answers.

Live Context connects to where work happens — captures decisions, patterns, and constraints as they're made — and triggers the right agent when something changes.

Confluence doc
Team meeting
Slack thread
Linear issue
Support ticket
GitHub PR
01 · Converge
03 · Serve
#sales-team

Pricing is $79/seat, annual only

Decisionratified
PR #342 · review

ADR-047 — data access via repository layer

Pattern4 approvals
AI Agents

Auth is OAuth2 + PKCE · 100 req/min

ConstraintCTO sign-off
Ticket #8847

Refund window is 14 days (Jan 8)

KnowledgePolicy Doc v4

Use cases

Every team has a blind spot. Live Context fills it.

Terminal
Tue Apr 1 9:41 AM
payment-service·main
Opus 4.6·4.2k tokens·$0.12
payment-service

[ How it works · 03 ]

Raw work in. Actionable knowledge out.

Three moves turn scattered conversations into knowledge your company — and its agents — can act on. Watch one Slack message make the whole trip.

Step 01

Converge

We connect to the tools where work already happens — Slack, GitHub, meetings, tickets — and capture raw knowledge at the moment it's created. No forms to fill, no wiki to garden.

SlackGitHubMeetConfluenceLinearIntercom+ more
#payments · Slack

“going with HMAC-SHA256 for Paddle signatures — CTO approved”

Architecture review · Meet

“…so we’re deferring DynamoDB to v2, Postgres for now…”

PR #342 · GitHub

“approved — repository pattern for all data access”

Ticket #8847 · Intercom

“why was my refund denied? I bought this 3 weeks ago”

Step 02

Classify

The layer cleans what it hears and types what matters — decisions, constraints, patterns — keeping the evidence and dropping the noise. What's left is signal, with sources.

evidence keptauthors keptnoise dropped
raw signaltyped knowledge
“HMAC-SHA256 for Paddle sigs”
ConstraintCTO sign-off

Paddle signatures must use HMAC-SHA256

“deferring DynamoDB to v2…”
Decisionratified

Postgres LISTEN/NOTIFY for v1 realtime

“anyone up for tacos? 🌮”
noise — filtered out

Step 03

Serve

Typed knowledge flows back to where it's needed: answering people inside their tools, loading agents over MCP, and triggering the right one when something changes. That's what makes it actionable.

people, in their toolsagents, over MCPtriggers, on change
Live Context

Paddle signatures must use HMAC-SHA256

ConstraintCTO sign-off
#paymentsSecurity review
Claude Code · MCP

Loaded before the agent writes the webhook handler

PR #342 · review

Cited in review before insecure code merges

#payments · Slack

Answers the next teammate who asks — with sources

[ Actionability · 04 ]

When knowledge changes, work happens.

Live Context doesn't wait to be asked. The moment a decision is ratified, it loads the new context into your agents over MCP — and the correction lands where your team is already working.

01 · A decision changes

DecisionDraft

Pricing is now annual-only

Proposed from #sales · awaiting review

Only ratified knowledge triggers action.

02 · The right surfaces act

Search Acme
# salesDeals, proposals, pricing questions
P
Priya Mehta2:14 PM

Sending Renewco the Q3 proposal now — quoting the monthly plan from the pricing deck.

github.com/acme/billing/pull/512
Add monthly plan picker to checkout#512
Opendana-fox wants to merge 2 commits into main from feat/monthly-plans
D
dana-foxcommented 20 minutes ago

Ready for review — adds the monthly billing option to the plan picker.

How it routes — one layer, four surfaces

team-shared · personal

Spaces & Notebooks

Team-shared Spaces hold what the team has settled. Personal Notebooks hold what you’re still working out. The boundary is explicit.

trust on every citation

Ask

Query the layer in plain language. Every citation arrives with its trust status — ratified, stale, or contested — so you know what the answer is standing on.

the layer, visible

Map

See how knowledge connects — which decisions feed which constraints, where the evidence lives, and where two truths collide.

every AI tool reads it

MCP

Claude Code, Cursor, claude.ai — every AI tool your team runs reads the same ratified context through one protocol. No copy-paste, no drift.

Converge your team’s knowledge.

Every source flowing into one ratified layer — read by every AI, trusted by every human. Stop digging. Start building on what your team already knows.

MCP-native · Claude Code · Cursor · claude.ai · any agent that can read