Can I compare them on my own corpus?

Yes, and this is the honest way to decide. The console can take any search from your history and re-run it against RAG-only retrieval, then show you both answers with both traces. The comparison run is logged underneath the original with the reason attached, so you build up a record rather than a vague impression.

Is the graph tier slower?

Yes. The plain search underneath still finishes at RAG speed; the extra time is the cross-document reach and writing the answer. Every response carries its own timings, so the number you plan around is the one measured on your corpus rather than one we quote — and if a particular surface has to feel instant, run that one against a RAG workspace and keep the graph tier where a second of thinking is fine.

What happens to documents I ingested before upgrading?

They stay indexed and searchable, and they are not retroactively extracted into the graph. If you want them in the graph, re-send them — ingest is an idempotent upsert on your own identifier, so a full re-POST of your corpus is safe and replaces rather than duplicates.

Can I go back to RAG?

A workspace stays on one tier for its lifetime, so the two stay comparable rather than drifting into a state nobody can reason about. For a single question you need nothing extra: the console re-runs any past search against RAG-only retrieval and shows you both answers with both traces. For a permanent side-by-side, run one workspace of each with the same content — the free tier holds a single workspace, so that wants Starter or above, which the fourteen-day trial gives you five of.

Comparison

Search your content, or search across it.

Both tiers are the same two calls over the same content. The graph tier changes how far a question reaches, and what the answer is allowed to say as a result — including things no single passage in your corpus contains. This page covers what that adds, what it costs, and when plain search is already enough.

See the retrieval engine Start free

At a glance

What each tier can answer.

RAG tier compared with the graph tier
	RAG tierfree	Graph tier
Retrieval
Searches within a document	Included	Included
Searches across documents	Not included	Included
Reaches content the query never worded	Not included	Included
Evidence is every document behind a claim	Not included	Included
Cited answers with enforced grounding	Included	Included
Diagnostics and timings	Included	Included
What ingest returns
Response	200 · indexed	202 · extracting + statusUrl
Searchable immediately	Included	Included
Graph populated	Not included	asynchronously
What the answer can say
Summarise a passage	Included	Included
Combine two passages the query matched	Included	Included
State a relationship no passage contains	Not included	Included
Name every document behind a claim	Not included	Included
Governance
Versioned ontology and drift	Not included	Included
Graph visualizer, entity quality	Not included	Included
Corpus health	Not included	Included

RAG tier

free

Searches within a document: Included
Searches across documents: Not included
Reaches content the query never worded: Not included
Evidence is every document behind a claim: Not included
Cited answers with enforced grounding: Included
Diagnostics and timings: Included
Response: 200 · indexed
Searchable immediately: Included
Graph populated: Not included
Summarise a passage: Included
Combine two passages the query matched: Included
State a relationship no passage contains: Not included
Name every document behind a claim: Not included
Versioned ontology and drift: Not included
Graph visualizer, entity quality: Not included
Corpus health: Not included

Graph tier

Searches within a document: Included
Searches across documents: Included
Reaches content the query never worded: Included
Evidence is every document behind a claim: Included
Cited answers with enforced grounding: Included
Diagnostics and timings: Included
Response: 202 · extracting + statusUrl
Searchable immediately: Included
Graph populated: asynchronously
Summarise a passage: Included
Combine two passages the query matched: Included
State a relationship no passage contains: Included
Name every document behind a claim: Included
Versioned ontology and drift: Included
Graph visualizer, entity quality: Included
Corpus health: Included

The difference, concretely

What a graph answer can say that a chunk cannot.

Both tiers find both of these passages. Only one of them can state what they add up to.

The question

Does finishing onboarding early actually reduce churn?

Two documents each hold half of it. Neither one contains the answer.

Cohorts that completed onboarding within seven days retained at 91% by day ninety, against 78% for cohorts that took longer.
Q3 Retention Review

The strongest predictor of a retained account is reaching first value quickly. Accounts that stall before activation raise tickets about scope, not about the product.
Support Themes 2026

The fact the graph holds

Onboarding completion —reduces→ Ninety-day churn (confidence 0.95, 3 sources)

The answer

Cohorts that complete onboarding within seven days show lower churn at ninety days [1], and the support record attributes the difference to reaching activation rather than to onboarding itself [3].

Both answers are accurate. Only one of them states the connection — that the second passage explains the first — because nothing in the corpus says so out loud. Plain search hands that inference to the reader; the graph tier makes it, and cites both documents for it. This example joins two documents. The one on the home page joins a billing record to a help article, which is the same mechanic across two applications rather than two files.

The trade

What the upgrade buys, and what each part costs.

Three things you get, and the price of each. Both columns are worth reading before you decide.

What you get

What it costs

Reach. Answers that pull in content the question never worded, and the full evidence behind every claim.

Latency. That reach is extra work and it happens before the answer is written. The plain search underneath still finishes at its usual speed.

A graph that persists, is queryable, and is rebuildable without paying to build it again.

Money. Model calls on every document at ingest, unbounded if anyone can point a firehose at it, which is why the tier needs a payment method.

Re-POSTing is safe and idempotent, so bringing your existing corpus in is a loop rather than a migration.

Documents ingested before you upgrade are not retroactively extracted. Upgrading changes what happens next, not what already happened.

Fit

Four cases where the free tier is already enough.

Worth checking against your own corpus first, so the graph is doing work it is genuinely good at rather than adding depth to a lookup.

Your questions are answered by single passages
If people ask "what is our refund window" and one document says, plain search already answers it. The graph is for the question one document cannot answer — keep it for the surface where those get asked.
Your content is one system, with one vocabulary
The graph earns its keep by joining entities across applications — the account in your billing system and the account in your ticket queue. With one source there is less to join, and the cross-source count that shows it working will read low. Worth revisiting the day a second system arrives, which is usually sooner than planned.
You need the lowest possible latency
A typeahead or an in-product search box that has to feel instant is a plain-search job. Run a RAG workspace there and the graph tier on the surface where a second of thinking is fine — they are the same two calls, so running both is one integration, not two.
You are still finding out whether retrieval helps at all
Ingest a few thousand documents and see what comes back. The graph adds reach across what you have, not content nobody wrote — so if the answers are thin because the material is thin, that is the more useful thing to learn first, and the refused questions are your list of what to write.

What the upgrade demonstrably buys, and the one thing we cannot yet measure.

The cross-document work contributes to 201 of 231 question runs on our evaluation corpus, so it is doing real work on most searches. What it demonstrably adds is the relationship no single passage states, the full evidence trail behind a claim, and a vocabulary your team can version and revert — all three checkable on your own content inside the trial.

What we cannot yet show is a percentage point of accuracy: on the question sets we can score, the graph has not been shown to answer a question that RAG-only retrieval missed. We are not going to dress that up. Buy it for the three things above, test them on your corpus, and if a measured accuracy lift is the only thing you need, tell us what you measured — that is the number we most want.

How we evaluate this

FAQ

Frequently asked

Compare them on your own content.

Start on the free tier, then run any query against both from the console when you want to see the difference.

no card required
first workspace is free
one POST to ingest

Talk to us Start free