Give every agent the right context and data, at the right time.
Sources
Your company’s brain
In your Git repo · Deterministic SQL engine · Validations · Access control
Agents
Agents don’t just write code. They optimize campaigns, resolve tickets and find insights in data.
Without it, they sound right and get it wrong.
The tribal knowledge: what things mean, how you do them, the workflows.
It is scattered across business apps and people’s heads. Without it, agents don’t know the business.
How the business is actually performing.
Built for humans and dashboards, not AI. It is ambiguous and duplicated, like 18 definitions of “active user.” So agents guess.
The AI enablement layer
Context and data definitions, together as code in your repo.
your repo
brain/
├── entities/
│ └── campaign/
│ ├── knowledge.md
│ └── schema.yml
├── glossary.md
├── policies.md
└── skills/
└── churn-investigation/SKILL.mdglossary.md
## activated user
A user who created a workspace and invited
three or more teammates. Trial signups are
never counted as activations.
used by: cost_per_activated_userYour context lives in your own Git repository. No vendor lock-in.
The build plugin keeps the brain healthy as your business and data evolve.
Agents pick named terms. The engine compiles them into the same SQL, every time.
The Lynk build plugin makes it a brain-building expert.
Slack a metric changesPRD a new releaseSnowflake a new table
Agents feedback loop
The brain stays correct as your business changes.
brain health · main
schema validations ✓ no conflicting definitions
evals ✓ 41/41 real tasks passing
freshness campaign package updated 2d ago
drift watch analytics.campaigns:
column `channel_group` added
→ PR #219 opened by the build agentIllustrative example.
The semantics engine validates that every entity, metric, feature and relationship exists exactly once.
Change the brain with confidence, as code.
Test changes against your agents before a new context version reaches production.
See the complete change log of every item in your company’s context.
The build agent and your teammates work on context changes side by side.
Changes stay visible in your ticketing system. The semantics live in your repository.
Every agent gets the brain, your data and the controls.
Brain context retrieval · Big-data analysis · Text-to-SQL · Data visualization
PII protection · Guardrails · Access control · Budget control
Lynk chat
Which campaigns ran over their cost-per-activated-user target this week?
Two: spring-sale-search (38% over) and q3-retargeting (31% over). Want the week-by-week breakdown?
context used: campaign package · glossary: activated user
One context hub for every person and every agent.
entities/campaign/schema.yml
entity: campaign
grain: one row per campaign
metrics:
spend: sum(spend_usd)
activated_users: count(distinct user_id)
cost_per_activated_user: spend / activated_users
