Goals

Q4 2026 objectives

We formed in September 2026. Our first objectives are temporary and focus on investigation. We will update them with specific delivery plans once we understand each problem and choose a direction.

Goal 1: Prove or disprove the context layer's value (Driver: Adam Bowker)

Before we invest in a context layer, we need evidence that it improves agent outcomes and solves a problem people care about.

What we'll ship:

• Interview 20 external users and five internal users, then synthesize what we learn
• Write a high-level request for comments (RFC) for the context layer
• Document how the context layer could unify Business Knowledge, the semantic layer, and PostHog's other context systems

Goal 2: Choose the right MCP architecture (Driver: Jake Ruth)

We need evidence that our Model Context Protocol (MCP) architecture offers the best experience for customers and agents.

What we'll ship:

• Prototype and compare promising approaches, including code execution and the current tool-based architecture
• Recommend an architecture based on what we learn
• If the current architecture remains the best option, update this objective to focus on improving it

Goal 3: Define metrics to measure the team's impact (Driver: Richard Solomou)

We need a set of metrics we can track to understand the team's impact on revenue, usage, or both.

What we'll ship:

• Define a small set of revenue and/or usage metrics that reflect the team's impact
• Document how each metric is calculated and establish a baseline
• Set up a dashboard and a regular review cadence to track progress against these metrics

Handbook

We make PostHog available wherever agents work and give those agents the organizational context they need to make better decisions. We own PostHog's Model Context Protocol (MCP) surface and are building a shared context layer that connects what a company knows across its tools and PostHog products.

What we own

• MCP – Make PostHog available wherever agents work
• Organizational context – Give agents shared company context so they can make better decisions

Working with other teams

Agents can see what happened in code or data, but they often cannot see why a decision was made, what a team is prioritizing, or what a company means by its internal language. We build the shared layer that gives agents this context without each product creating its own memory system.

Come to us if you:

• Are building an integration with the PostHog MCP server or store
• Want an agent to read from or contribute to organizational context
• Want to connect a source such as Slack, GitHub, conversations, or your semantic layer
• Need help with privacy, access control, or safe context sharing between agent surfaces

Feature ownership

You can find out more about the features we own.