PromptQL Review (2026): Pricing, Features & Honest Verdict

PromptQL is an AI data agent from Hasura that you point at your databases and tools, then ask questions in plain English and build workflows as a team. The 2026 version adds multiplayer threads and a shared team brain. It is genuinely useful for data-heavy teams, but the consumption pricing takes nerve. Best for data and ops teams. Price: pay-as-you-go from about $0.14 per task unit, free credits included. Rating: 7.5/10.

What is PromptQL?

PromptQL is Hasura’s AI analyst. You connect it to your databases, warehouses, and business tools, then ask questions in normal language and get answers backed by your actual data. It launched in mid-2025 as what Hasura called the spiritual successor to GraphQL, aimed at letting anyone query company data without writing SQL or waiting on the data team.

The 2026 relaunch changed the pitch. PromptQL is now sold as multiplayer AI, and the tagline literally says it replaces Slack. The idea is shared threads where teammates tag each other to review and correct answers, plus a shared wiki that captures your team’s definitions and edge cases automatically. Hasura is an established company with a real engineering track record, so this is not a weekend project. It is a serious attempt to build a team layer on top of AI data analysis.

To picture how it works in practice, imagine a revenue ops lead asking why enterprise renewals dipped in Q2. PromptQL plans out the query, pulls from the CRM and the billing warehouse, and comes back with a segmented answer plus the steps it took to get there. A teammate spots that the definition of “enterprise” was wrong, fixes it in the thread, and every future answer inherits that correction. That loop, ask then correct then remember, is the whole product in one sentence. It is closer to onboarding a junior analyst than to using a search box.

What Are PromptQL’s Key Features?

Plain-language queries against your real data

The core function is simple to describe. You connect a data source, then ask something like “which customers churned last quarter and what did they have in common.” PromptQL plans the query, runs it against your data, and returns an answer with the reasoning shown. It supports 30 plus models under the hood, so you are not locked to one provider. This is the part that works well and justifies the tool existing at all.

Multiplayer threads with a shared team brain

This is the 2026 headline feature. Instead of everyone running private ChatGPT sessions, PromptQL puts analysis in shared threads. You tag a teammate to review an answer, they correct a definition, and that correction sticks. Over time the shared wiki learns what your team means by “active user” or “qualified lead” so the AI stops giving inconsistent answers to the same question. If your team argues about metrics in Slack all day, you can see why Hasura framed it as a Slack replacement.

Automatic context capture

PromptQL claims to build institutional knowledge without documentation overhead. As people ask questions and fix answers, the system records the domain knowledge behind those fixes. That means new hires can ask the tool instead of pinging a senior analyst. Whether it captures context as cleanly as advertised depends on your data being reasonably structured, but the concept is sound and it is a real differentiator versus a plain chatbot.

Workflows, not just answers

PromptQL is not only a question box. You can turn a repeated analysis into a workflow that runs on demand or on a schedule, so the weekly churn report or the daily pipeline summary generates itself. This is where the OLU pricing starts to make sense, because a saved workflow is cheaper and more consistent than a person rebuilding the same query every Monday. It also nudges the tool from a novelty into something that lives in your actual operating cadence.

Bring your own LLM and cloud

On the Enterprise tier you can bring your own model keys from Anthropic, AWS Bedrock, or Google Vertex, and run inside your own VPC or cloud. For teams with sensitive data that cannot leave their infrastructure, this matters a lot. It is the difference between a tool security will approve and one they will block. Hasura also adds audit trails and fine-grained permissions here, so you can control who can query which tables. That governance layer is what separates PromptQL from pointing a generic chatbot at a database export and hoping for the best.

One honest caveat runs through all of these features. AI on top of real data can still be confidently wrong, and a plausible-looking chart with a bad join underneath it is more dangerous than an obvious error. PromptQL showing its reasoning helps, because you can audit the steps, but you still need someone who can sanity-check the output before a number lands in a board deck. Treat it as a fast first draft of an analysis, not as an oracle.

How Much Does PromptQL Cost?

PromptQL uses pay-as-you-go consumption pricing, and this is where opinions split. There is no monthly seat fee. Instead you pay per OLU, which Hasura defines as a normalized unit of tokens, at an introductory rate of about $0.14 (standard rate $0.20). A simple data task runs under $0.28, a complex report around $1.40, and a deep investigation around $5.60. New projects get up to $100 in free credits plus $70 for every teammate who joins, and no credit card is needed to start.

Compare that to a per-seat analyst tool and the math gets interesting. If your team runs a few dozen queries a week, you might spend less than a single ChatGPT Team seat. If ten people are running deep investigations all day, the bill can climb fast and it is hard to predict in advance. That unpredictability is the real cost. Enterprise pricing is custom and adds security, audit trails, SSO, and BYOC. The consumption model is fair, but finance teams that like fixed numbers will flinch.

There is a smart wrinkle worth knowing. Hasura lets you route work to cheaper models and quotes big savings against a top-tier baseline, so a task that costs a dollar on a premium model can cost cents on a lighter one. For routine questions that is real money saved at scale. The flip side is that you now have to think about model choice per task, which is one more variable most teams did not sign up to manage. Set per-user quotas early if you roll this out widely, or a single curious analyst can burn through credits exploring.

Who is PromptQL Best For?

Use PromptQL if you are a data-heavy team drowning in ad hoc data questions, and you want non-analysts to self-serve without SQL. It is a strong fit for ops, revenue, and product teams inside companies that already have their data in a warehouse. The multiplayer angle pays off most when several people need the same answers to mean the same thing.

Skip PromptQL if you are a solo founder or a small team without a real database yet. The tool assumes you have structured data worth querying. If your “data” is three spreadsheets, a plain AI chat is cheaper and simpler. Also skip it if unpredictable monthly bills would give your finance lead a heart attack, because usage-based pricing is exactly that.

Best PromptQL Alternatives

Julius AI: A more consumer-friendly AI data analyst that works well with spreadsheets and CSVs. Easier to start with, weaker on connecting to production warehouses and team context. Plans from around $20/mo.

Teable: An Airtable-style AI database that turns your data into apps and workflows. Better if you want to store and structure data too, not just query it. Has a free tier and paid plans from about $12/user/mo.

ChatGPT Team with connectors: The generalist option. Cheaper and more flexible for mixed work, but it lacks the built-in data governance, shared metric definitions, and warehouse-grade querying that PromptQL is built around. From $25/user/mo.

Final Verdict: Is PromptQL Worth It?

PromptQL is one of the more thoughtful takes on AI for data teams we have found this year. The core querying works, and the multiplayer layer solves a real problem: teams that can never agree on what a metric means. Hasura clearly understands data infrastructure, and it shows in the BYOC and governance features. This is not a thin wrapper.

The catch is the pricing model and the maturity of the “replaces Slack” claim, which is more marketing than reality today. If you have a warehouse, a team, and recurring data questions, grab the free credits and test it against your real workload before you commit. If you are small, early, or allergic to variable billing, wait. For the right team it is a 7.5. For the wrong one it is an expensive experiment.