Use your Knowledge Graph inside Claude

Using a business Knowledge Graph inside a Claude session means the entities, metrics, definitions, and decisions a company already maintains in SIGNLD are available to Claude while someone works, instead of being explained by hand in a prompt or looked up separately in another tool.

What this actually does

A Knowledge Graph, in SIGNLD's sense, is the structured map of a company's entities (customers, vendors, products, teams), the metrics that matter to the business, the definitions those metrics depend on, the rules that govern how they're calculated, and the decisions already made about them. SIGNLD builds this graph from a company's connected systems and keeps it current.

The SIGNLD extension for Claude is what makes that graph available inside a Claude session. The SIGNLD extension for Claude is built and maintained by SIGNLD using Anthropic's publicly available extension interfaces. SIGNLD is not affiliated with, sponsored by, or endorsed by Anthropic.

In practice, this means someone working in Claude doesn't need to re-explain what "qualified pipeline" means at their company, or which system owns the customer count, because that context is already sitting in the Knowledge Graph and is available in the session. The extension surfaces it; SIGNLD builds and holds it.

This is a different kind of thing than pasting a spreadsheet or a paragraph of background into a chat. A pasted paragraph is a snapshot, written by one person, that goes stale the moment the underlying number changes. The Knowledge Graph is maintained continuously from the source systems it's built from, so what's available in the session reflects what's currently true in SIGNLD, not what someone remembered to copy in.

Consider the difference in a concrete case. Say a company defines "active customer" a specific way, counting only accounts with usage in the last 30 days, excluding trial accounts, and rolling up multiple subsidiary accounts under one parent. That definition took time to settle on and it's already encoded in SIGNLD's Knowledge Graph. Without the extension, a person asking Claude about active customers either has to restate that whole definition every time or risks getting an answer built on a generic, unstated assumption about what "active" means. With the Knowledge Graph available, that specific definition is what the session works from.

What it does not do

The extension does not give Claude write access to any system the Knowledge Graph is built from. It is a read layer on top of context SIGNLD already holds. It doesn't create new data, and it doesn't change anything in a connected system.

It does not describe or rely on any specific Claude feature, plan, limit, or model name, because those belong to Anthropic and change on their own schedule. Anything specific to Claude itself is documented at docs.anthropic.com.

It also doesn't build the Knowledge Graph itself. If a company hasn't connected its systems to SIGNLD, there's no graph to bring into a Claude session in the first place. The extension is the bridge between an existing graph and a Claude session, not a way to generate one from scratch inside the conversation.

This scoping matters for a reason beyond neatness. It's tempting to describe an integration like this in terms of everything it might someday do, but being specific about the boundary is what keeps the read-only guarantee meaningful. A graph that can be read but never written through the extension is a graph a company can trust to stay intact, regardless of what happens inside any individual Claude session.

How it works, step by step

First, SIGNLD builds and maintains the Knowledge Graph from a company's connected systems: entities, metrics, definitions, rules, and decisions.

Second, a person connects the SIGNLD extension for Claude to their SIGNLD account, using the permission flow Claude's own extension interface provides.

Third, inside a session, the relevant parts of the Knowledge Graph become available as context Claude can reference, without the user typing definitions or background information in manually.

Fourth, where an answer touches a specific entity, metric, or decision from the graph, that reference stays traceable back to the record in SIGNLD, and from there back to the source system it came from originally.

Fifth, everything here stays read-only against the systems the graph is built from. The extension surfaces context; it does not act on the underlying systems.

None of these steps depend on the user knowing which system a piece of context originally came from. The Knowledge Graph already did the work of resolving an entity across systems, so a person asking about a customer doesn't need to know whether that customer's record started in a CRM, a billing system, or a support tool. What they get is the resolved view, with the underlying trail preserved for whenever it needs to be checked.

What changes for the person asking the question

Without a shared Knowledge Graph inside the session, a person working in Claude on a business question has to supply company context themselves, which is slow, easy to get wrong, and has to be redone every session because the conversation doesn't carry that context forward on its own.

With the Knowledge Graph available, that context doesn't need to be reconstructed. A person can ask a question that touches a specific metric or entity, and the definition, ownership, and history already established in SIGNLD is present rather than assumed. That also means two people on the same team, each in their own Claude session, are drawing from the same graph rather than two different mental models of the business, which matters when the answers they get feed into a shared decision.

It changes what counts as a checkable answer, too. A response that names a specific entity or metric from the graph can be traced to where that entity or metric lives and how it was defined, which is different from a response that sounds reasonable but has no path back to anything concrete.

There's a broader shift this points at, even at the level of a single session. Most of the effort in getting a useful answer from a general assistant on a business topic goes into supplying context the assistant doesn't have on its own: what a term means at this specific company, who owns a number, what's already been decided about how it's tracked. A Knowledge Graph made available inside the session removes most of that setup cost, because the context isn't being reconstructed from memory each time, it's being read from the same place the rest of the company already relies on.

Permissions and data access

The SIGNLD extension for Claude is built and maintained by SIGNLD using Anthropic's publicly available extension interfaces. SIGNLD is not affiliated with, sponsored by, or endorsed by Anthropic.

Everything the extension surfaces inside a Claude session stays read-only against SIGNLD's connected systems, and stays traceable back to the specific record it came from. Connecting the extension doesn't grant SIGNLD new access to any system, and it doesn't let a user see Knowledge Graph context they didn't already have access to inside SIGNLD.

For SIGNLD's broader security posture, see /security. For how the Knowledge Graph is structured, see /concepts#knowledge-graph.

Key takeaways

A Knowledge Graph, in SIGNLD's sense, is the structured map of a company's entities (customers, vendors, products, teams), the metrics that matter to the business, the definitions those metrics depend on, the rules that govern how they're calculated, and the decisions already made about them.
The extension does not give Claude write access to any system the Knowledge Graph is built from.
First, SIGNLD builds and maintains the Knowledge Graph from a company's connected systems: entities, metrics, definitions, rules, and decisions.
The SIGNLD extension for Claude is built and maintained by SIGNLD using Anthropic's publicly available extension interfaces.
Yes, this describes what the extension actually enables.

FAQ

Do I need to explain my business to Claude for this to work?

No, that's the point. The Knowledge Graph SIGNLD already built from your connected systems is available in the session through the extension, so definitions, entities, and metrics don't need to be typed in by hand each time.

Is the Knowledge Graph a separate copy inside Claude?

No. The extension reads from the same Knowledge Graph SIGNLD maintains. There isn't a second version of it living inside Claude that could drift out of date on its own.

Can Claude change anything in my connected systems through this?

No. The extension is read-only. It surfaces context that already exists in the Knowledge Graph; it doesn't write to, edit, or delete anything in the systems that graph is built from.

What if my company hasn't connected many systems to SIGNLD yet?

The Knowledge Graph reflects whatever's connected. A thinner set of connections means a thinner graph available in the session, but it grows as more systems are connected, without needing anything different from the extension itself.

Is this the same as the SIGNLD extension for Claude?

Yes, this describes what the extension actually enables. For a fuller look at the extension itself, including setup and permissions, see the SIGNLD extension for Claude.

Try SIGNLD free or see how it works.

Claude and Anthropic are trademarks of Anthropic. Used descriptively only.

Use your business Knowledge Graph inside a Claude session

In this article