ClariLayer icon

ClariLayer

ClariLayer is a context layer for AI data agents that connects to Claude Code, Cursor, Codex, or claude.ai. It keeps definitions durable across sessions and checks them against source evidence.

ClariLayer

What ClariLayer does

ClariLayer is a context layer for AI data agents. It connects to Claude Code, Cursor, Codex, or claude.ai and helps an agent keep durable data context across sessions instead of starting from zero each time.

The product’s core job is to preserve and check definitions such as tables, joins, and business metrics against source evidence. In the personal MCP path, ClariLayer reconciles what the agent believes with warehouse or HubSpot evidence, and a mismatch becomes a caveat rather than an unverified assertion.

The site also separates a free individual product from a team-oriented governed product. Personal use is free and unmetered, while the Governed Context Edge for teams is in private pilot with early access available by request.

Core capabilities

Agent connections

Connects to claude.ai over OAuth, or to Claude Code, Cursor, and Codex through MCP, so the agent can recall context inside the workflow it already uses.

Bootstrap from existing context

Starts from five source kinds — SQL, dbt, CLAUDE.md or notes, a data dictionary, and semantic models — so existing context can be imported instead of rebuilt manually.

Recall and remember

Lets the agent recall saved context mid-task and remember corrections across sessions, reducing the need to re-explain table names, joins, and definitions.

Reconcile against evidence

Checks a saved definition against source evidence and marks disagreements as caveats instead of silent assertions, which is the central distinction from a notes file.

Completion receipt

Records a durable completion receipt before the agent says done, so the session can leave behind a visible context declaration or explicit nothing-to-update outcome.

Where ClariLayer fits

  • Maintain personal analytics context

    A solo analyst can keep a stable definition for metrics like active customers or net revenue, so the same wording does not need to be retyped into every new chat.

  • Check metric definitions against warehouse evidence

    When a saved warehouse definition is reconciled against actual evidence, the agent can surface null handling, join-path, or row-count caveats before returning a result.

  • Bridge from solo work to team governance

    A team that is beginning to depend on shared definitions can start with the free personal layer now, then move to the governed team offering once the context becomes collaborative infrastructure.

  • Validate HubSpot CRM definitions

    A HubSpot workflow can store a CRM contract and reconcile it against row-free crm_evidence, giving users a way to check agreed property definitions without sending records or credentials to ClariLayer.

  • Upgrade from notes to checked context

    A user moving from a notes-only CLAUDE.md workflow can bootstrap structured context from existing files, then let later corrections persist across sessions instead of staying as unverified text.

Pros and Cons

Pros

  • Works with Claude Code, Cursor, Codex, and claude.ai.
  • Can bootstrap from several existing knowledge sources instead of requiring a blank setup.
  • Treats mismatches as caveats, which helps users spot stale or conflicting definitions.
  • Free and unmetered for individual use, with no card mentioned on the pricing page.
  • Keeps context durable across sessions so teams and individuals can avoid re-explaining the same data.

Cons

  • The team offering is not publicly available yet and is in private pilot.
  • HubSpot reconciliation is the only CRM path described as reconcilable in the current contract; Salesforce is recall-only after CRM access passes reconcile.
  • For personal use, ClariLayer does not hold warehouse or CRM credentials or run SQL server-side, so the agent still depends on its own connected data access.

FAQ

How do I connect ClariLayer to my AI agent?

ClariLayer connects to claude.ai over OAuth, or to Claude Code, Cursor, and Codex through an MCP connection. The source also says the personal path is free and unmetered.

Is ClariLayer for solo use or teams?

The product is built for individuals who want their AI to remember and check data context across sessions. The site describes a free single-player core and a separate governed team product in private pilot.

What does reconcile do?

ClariLayer supports recall, remember, bootstrap, and reconcile. In the documented reconcile flow, it checks a saved definition against warehouse actual_sample evidence or HubSpot CRM evidence and turns mismatches into caveats.

What kinds of source material can it use?

The source says ClariLayer can bootstrap from SQL, dbt, CLAUDE.md or notes, a data dictionary or codebook, and semantic models. It also says the MCP path does not require ClariLayer to hold warehouse or CRM credentials.

Does ClariLayer have a team plan?

Yes, but the team offering is separate. The Governed Context Edge for teams is in private pilot, with early access requested directly from ClariLayer; the public pricing page says there is no public price yet.

Quick Facts

Category
AI data agent context layer
Primary users
Individual analysts and AI data-agent users
Platforms
claude.ai, Claude Code, Cursor, Codex
Connection method
OAuth for claude.ai; MCP for coding agents
Pricing
Free and unmetered for personal use; team product in private pilot
Source domain
clarilayer.com