Heplon

What Heplon gives your coding agent

Heplon puts your platform's knowledge and its live state inside your coding agent. You connect Heplon to your agent once. After that, the agent can reason about how your systems work, not only about the code in front of it.

The problem it solves

A coding agent knows your repository. It does not know your platform. It does not know which services yours depends on, who owns them, what your deployment and observability standards are, or what policies apply to the code it writes. So its output looks reasonable in isolation and then fails to fit the environment it runs in.

Heplon closes that gap. It gives the agent a grounded view of your platform, so the agent can produce work that fits.

Where the knowledge comes from

Heplon builds a knowledge graph by ingesting your systems: source control, service catalogs, and more. The graph models services, their dependencies, ownership, policies, and the evidence behind each fact. See the platform graph.

Heplon also reads live state. For example, a Kubernetes connector lets the agent read current cluster resources. Live functions are read-only, so the agent can look but cannot change an upstream system.

So the agent works from two sources at once: the graph, which is the stable model of your platform, and live functions, which are the current state.

Two tools

Heplon exposes two MCP tools. Your agent calls them for you.

  • ask answers a question about your platform. For example, what runs billing, or what a service depends on.
  • guide briefs the agent before it does a piece of work. You give it an intent, such as "write a service that reads from the orders datastore". It returns the coding standards, related and dependency services, observability standards, and policies that apply. The agent uses that briefing to produce work that fits your platform.

For the exact inputs and outputs, see the MCP tools reference.

What you can do with it

See also