PromptQL
Multi-player AI workspace allows the team's context to automatically accumulate rather than decay in daily work. Everyone corrects it once and everyone benefits.
In-depth Report
-
PromptQL is a "multiplayer AI workspace" produced by Hasura. It uses shared threads to replace Slack-style fragmented communication, allowing the team's context to automatically accumulate rather than decay in daily work. It was launched on Product Hunt on July 23, 2026 with the positioning of "Multiplayer AI that replaces Slack" and received 246 votes, ranking third on the day. The key difference between PromptQL and ordinary AI chat tools is that it is not just one person asking the AI, but the entire team shares a self-learning AI brain - whoever corrects once, everyone benefits. It has been verified within the 70-person team, and monthly activity data shows that wiki contributions have increased from less than 50 entries per day to over 300 entries. The core selling point of PromptQL is not "faster answers", but "fewer duplicate instructions and less information loss".
-
PromptQL comes from Hasura, a star company in the GraphQL ecosystem. Founders Tanmai Gopal and Rajoshi Ghosh founded Hasura in Bangalore in 2017, and open sourced the Hasura GraphQL Engine in 2018. To date, it has accumulated more than 400 million downloads and 31,000 GitHub Stars. The company has raised a total of $136.5 million in funding, with a $100 million Series C round led by Greenoaks in 2022, valuing it at $1 billion. Investors include Lightspeed, Nexus Venture Partners, Vertex Ventures. Hasura's main business has always been "enabling applications to access data faster" - generating APIs with zero code through the GraphQL engine. But Tanmai has expressed a judgment many times in interviews: "What needs to access data now (no longer applications), but AI. AI is accessing data for users." This judgment directly gave birth to PromptQL. In Tanmai's words, PromptQL is "the natural continuation of GraphQL in the AI era." PromptQL was launched as an independent product in June 2025. It was initially positioned as an "AI data access agent", which uses natural language to replace SQL/API to query enterprise data. By mid-2026, the product direction has evolved significantly—expanding from pure data layer tools to a complete multi-player AI workspace. The "About" page of Hasura's official website has been redirected to promptql.io, indicating the priority of this transformation at the company level. The founders publicly stated that the company has fully migrated to PromptQL internally (70-person team) starting in February 2026, replacing the Slack used in the past. According to Tanmai's description on X, this experiment "triggered team-wide AI psychosis" - the shared AI began to aggregate information from all channels, warehouses and project boards - but it proved that the product had truly become "Slack for the AI era."Hasura currently has about 170 employees. In 2025, Suku Krishnaraj Chettiar (former CMO of Sumo Logic, who took it from $10 million to $300 million ARR) joins as COO. PromptQL has been deployed in a $900/hour AI engineer consulting model at Fortune 500 clients, with seven-figure contracts signed within months. However, the company has not raised funds since Series C in 2022, and has not released any ARR data, indicating that PromptQL’s business model may still be in the exploratory stage.
-
The functional system of PromptQL can be broken down into three layers: data connection layer, reasoning execution layer and team collaboration layer. Data connection layer: PromptQL can connect to databases (Snowflake, PostgreSQL, etc.), SaaS tools (Salesforce, Slack, Notion, etc.) APIs (via MCP protocol) and files (Google Docs, spreadsheets, etc.). The connection process does not require the establishment of ETL pipelines or data replication. It uses a virtual query layer and the data remains in place. This is inherited from the technical architecture of Hasura DDN (Data Delivery Network) - a set of metadata layers are used to describe the meaning of data, and AI queries through it without directly touching the underlying storage. The official promise is "zero data copying" and "data never leaves your borders." Speculative execution layer: This is the most impressive part of PromptQL and the fundamental difference from normal text-to-SQL or RAG solutions. After the user asks a question, PromptQL does not directly generate an answer. Instead, it first generates a readable and editable query plan (Query Plan), explaining step by step what the AI is going to do - which sources to fetch data from, what calculations to make, and what assumptions to use. Only after the user reviews and confirms (or modifies) the plan does it enter the execution phase. The execution results are stored as artifacts (tables, charts, text), independent of the model context, so it can handle data volumes that far exceed a single context window. This "plan first, execute later" architecture solves LLM's most vexing problem: making up answers. PromptQL gives a Reliability Score for each answer, generated based on query complexity, data quality, and model confidence. When the score is low, the AI will make it clear that it is unsure rather than forcing output. Citing data from the Berkeley Data Agent Benchmark, the current best score of the cutting-edge model is only 43% (Opus 4.6), and the vast majority of text-to-SQL solutions are close to 0%. The founding team of PromptQL believes that this gap is the crux of the "accuracy problem." Team collaboration layer: This is a core new value launching in July 2026. PromptQL's thread is "shared by multiple people": when you ask a question, the AI's reasoning process and intermediate results are visible to everyone in the team. Colleagues can jump in to add, correct, or like. Correction once becomes shared knowledge (Wiki entry, Skillor semantic model changes), no one will need to repeat the explanation in the future. Essentially, PromptQL turns "maintaining context" from an additional burden into a by-product of normal workflow. At the UX level, the product looks like a chat interface with an AI assistant, but supports Markdown rendering, tables, and charts. Search results are displayed in Artifact cards. The interaction design pays attention to the details - when the AI wants to learn something new, it will automatically pop up the "PromptQL wants to learn" prompt, and the user will directly write it to the Wiki after confirmation. The entire experience feels more like collaborating with a smart colleague than operating a tool.
-
PromptQL currently offers a free version (3 seats + limited query quota) and a paid Pro version (complete data warehouse integration, wiki generation functions). There are customized solutions for corporate customers. From a business model perspective, Hasura has adopted multiple strategies: first, SaaS subscription (billed by seat + query volume); second, high-end AI engineer consulting services ($900/hour), helping Fortune 500 customers complete PromptQL deployment customization within two weeks; third, Token recharge (supporting multiple models such as Fable, Sol, Kimi-K3, etc.), and a $1,000 Token credit was given to Product Hunt users when the product was released. This multi-pronged strategy shows that PromptQL is still in the stage of exploring the optimal pricing model. By comparison, competitive Glean pricing for enterprises starts at tens of thousands of dollars per year. If PromptQL can build a reputation among small and medium-sized enterprises with the free version, it is not uncommon to move up to the enterprise version. However, a signal worth paying attention to is that Hasura's main business model (GraphQL Engine cloud version + self-hosted version) may have internal competition with PromptQL's pricing system. If the team is already using Hasura GraphQL Engine for the API layer, why pay extra for an AI layer that does the same thing with natural language? This cross-positioning question has yet to be answered head-on by Hasura.
-
PromptQL has a 5.0/5.0 rating (7 reviews) on Product Hunt, with overall positive user feedback. Hashnode’s last in-depth review gave it a 3.5/5, saying the product is real and effective but feels abrasive. Positive comments focused on the difference point of "shared context": - Shahidh K Muhammed, a director-level engineering leader said: "I used to spend my time switching tools - Grafana, Slack, AWS, Snowflake, GitHub, Linear, endlessly. The real work is not hard, but the friction between the work. PromptQL keeps me in a state of flow." - Marketing Manager Lili Riahi said: "No matter what the task, my starting point is opening a new thread in PromptQL. It's not an exaggeration. It completely changed the way I work." - Anushrut Gupta, head of AI, said: "I no longer maintain a TODO list. Not because there are fewer things to do, but because the unit of work is no longer a task in the queue - it is an already running thread." Negative feedback and points of criticism: - Product positioning drift: In the early days, it was "AI data access agent", now it is "multiplayer AI workspace", users may be confused about what they are buying. - Initial setup requires investment: connecting multiple data sources and teaching it business context takes time and is not ready out of the box - Propagation lag problem: When encountering a conflict between two data sources, PromptQL will not automatically resolve it. Instead, it will mark the conflict and ask the team. This is rigorous but slows down the pace. - The free version has limited functions: it cannot fully test the core selling points (multi-source data integration + Wiki generation), and the Pro version is required to evaluate the true value
-
Attention from the media and industry analysts increased significantly following the product relaunch in July 2026. RuntimeWire believes in the report that the difference of PromptQL is "treating the entire workspace as an AI that can generate task-specific Agents on demand", not just "AI-assisted team chat." But the article also questions: Slack, Microsoft Teams and Microsoft Copilot are already embedding AI, and user habits and platform network effects are huge obstacles that PromptQL needs to overcome. Grayscale VC's analysis article gave high praise from the perspective of technical architecture and believed that PromptQL's "plan-execution separation architecture" is significantly better than the three mainstream solutions of MCP tool orchestration, multi-agent orchestration and text-to-SQL in the three dimensions of predictability, controllability and user-friendliness. The core argument is that LLM is good at planning but not good at execution, and PromptQL makes up for this shortcoming with deterministic execution. Crows Eye Intelligence Dossier's overall analysis of Hasura mentioned: The community is divided on Hasura's transformation from GraphQL to AI - the architectural changes of V3/DDN have made many old users feel abandoned, and the increase in cloud version pricing has also made some users dissatisfied. Whether PromptQL's high level of attention can compensate for the emotional drain from the core GraphQL community is an open question. VentureBeat's report quoted Tanmai's core judgment: "The biggest problem is not the inability to answer questions accurately. The biggest problem is that AI pretends to be accurate even when it is not accurate. AI is confidently making mistakes. That is the problem." This statement accurately captures the problem PromptQL wants to solve - reliability, not speed or breadth.
-
The most significant risk faced by PromptQL is "transformation pain." The trust in the GraphQL community that Hasura has spent nearly 6 years building—31,000 GitHub Stars, 400 million downloads, 30+ high-profile enterprise customers—is now reinvested in a new product that has yet to have proven market fit. There are already voices of dissatisfaction in the community: the unusual changes in V3/DDN have confused many old users, and the increase in cloud version pricing also makes people miss the V2 era. If PromptQL fails to achieve expected revenue, Hasura may face the embarrassment of losing both its GraphQL base and its foothold in the AI field. In terms of privacy and security, PromptQL promises "data never leaves your boundaries" and "zero data copying." The system architecture does implement a virtual query layer and attribute-level policy execution. But in an X discussion, a user asked whether AI Agents could accidentally expose sensitive APIs when creating new connectors, and the founders responded that each integration is subject to the owner's access controls and developers must explicitly authorize it. There is currently only an official description of this mechanism, and there is still a lack of independent third-party security audit reports. In the competitive landscape, PromptQL faces pressure from multiple directions: horizontally, there are Glean (knowledge management and AI search) and Notion AI (document + AI); vertically, there is Microsoft Copilot for Teams (embedding into the existing office ecosystem); and at the technical layer, there are LangChain/LlamaIndex (Agent framework) and Defog (text-to-SQL). Each competing product has its own moat. Although PromptQL's "shared context" is a unique value proposition, whether it can tear a hole in the existing collaboration tool ecosystem depends on whether it can make enterprise teams willing to migrate collectively - and this is precisely the most difficult thing. Product positioning drift is also a risk signal. The official website has changed from an "AI data agent" to a "multi-player AI workspace", which the founder calls an "accidentally discovered product". This kind of claim represents a real demand signal, but it may also mean that the product team has not fully locked in the target market and user profile.
-
The most suitable teams for PromptQL are organizations with many data sources, frequent cross-department collaboration, and those who are fed up with "answering the same questions repeatedly on different occasions". Specifically: Suitable for: Analytical teams in operation-intensive enterprises (retail, logistics, finance) who need to make regular reports and data-driven decisions across Snowflake, Salesforce, and internal systems; small and medium-sized technology teams who are tired of Slack information overload and are willing to try new working models with AI threads as the core; teams with cross-regional collaboration needs that want to unify business contexts scattered everywhere. Not suitable for: small teams with a thin data stack (a Google Sheet is enough); enterprises with extreme data security requirements and unwillingness to rely on third-party SaaS; teams that have invested heavily in Glean or other knowledge management tools. In terms of alternatives, if you just need text-to-SQL, the AI query function of Defog or Supabase is more lightweight; if it is mainly knowledge management, Glean and Notion AI are more mature; if you just want AI-assisted chat, Microsoft Copilot for Teams or Slack AI are more painless (at least no team migration is required). The uniqueness of PromptQL is that it integrates these requirements, but the cost of integration is higher learning costs and migration costs.
-
PromptQL is a product with a bold vision - replacing fragmented team communication with shared AI context - but its success does not depend on whether the product itself is good or not, but on whether it can enable the team to collectively migrate to a new working paradigm. In terms of technical architecture, "plan-execution separation" and "self-correcting context layer" represent the cutting-edge level in this field. At the market level, Hasura's GraphQL roots provide it with technical credibility and existing corporate relationships, but also introduces the risk of distraction and unclear positioning. For teams willing to try new working models, PromptQL is worth a try; for companies that need certainty about ROI, it is recommended to wait until it runs for a few quarters.
User Reviews
-
RMurphyX98—After using it for three days, I no longer want to go back to Slack. The idea of sharing context is too painful. -
KCampbell_7784—PH gave a 1,000 yuan token quota, and the team was invited to give it a try. How should I put it? I was confused for the first half hour. After I got used to it, I realized that AI can collaborate with a team in this way. -
MMendoza_Max—PromptQL is really redefining team collaboration. There is no need to repeatedly transfer files, just share a thread, and the AI agents can each do their own work. This is a great experience. -
Msدینااحمدی_2024—Correct it once and everyone benefits from that. This mechanism of “turning what you learn into shared skills” is so right. There is no need for everyone to teach AI individually. -
RMorales_2021—The permissions aspect is quite solid. The same AI will see different data for different people. Finance can only look at finance, and sales can only look at sales. This allows companies to use it with confidence. -
Scott_Peterson_2024—After using PromptQL, I no longer have to jump back and forth between various tools every day. I can just ask for data from Salesforce, Marketo, and Outreach directly. In the past, just replying to questions such as "How is the event registration?" and "What information should I send to this customer?" took up most of my energy. Now they ask AI directly, and I finally don't have to be an information relay station. -
JBakerZ—@nl_kapur is so right, all cross-department context is stored in one place, no need to look through chat history anymore! -
nODErUNNERrEED—It has been two months since our team migrated from Slack to PromptQL. The biggest feeling is that the context is not lost. In the past, when changing a project, we had to go through the chat history all over again. Now the AI directly knows what we have discussed before. It feels so good. -
mobp4kqkvf—There are so many connectors, Snowflake, Postgres, MongoDB, BigQuery are all supported, and even the MCP protocol is supported. It feels like Hasura’s accumulation in data connection is fully utilized. -
CTurner_66—It's not easy to move a team to a new tool en masse, but it's the right direction. Humans are responsible for context, AI is responsible for execution, and the division of labor is clear. -
KennethWood_Plus502—It was indeed painful to migrate to Slack, but once you get used to it, you can’t go back. The most obvious change is that we have a lot fewer meetings, because AI can directly answer "What is Alex busy with" and "Where is this project stuck?", and there is no need to specially invite people to meetings to ask questions. -
Emily_RamosQ62—The price is based on consumption, not per person, which is a good point. If the team has more people, we are not afraid of doubling the charge. -
ticklishladybug283—Reviews say that its planning-execution separation architecture is the most impressive part, and this is indeed the case in my actual testing. After asking a question, AI will first show you how it wants to check the data and what it wants to calculate, and then you confirm it and then execute it. This kind of transparency is not seen in other AI tools. -
JSullivan_X98—The disadvantages are also quite obvious. The initial setup takes time, and you have to teach it to understand your business. Not a ready-to-use product out of the box. And once misinformation accumulates in a wiki, it can be scary to spread it. -
Jesse_Murphy45—The Hasura team is really powerful. They can achieve 31k stars for GraphQL, and now they also have a set of AI workspaces. PromptQL's "plan first, then execute" model solves the problem of AI making up, which is really reliable. -
Joseph_Morales_Plus225—Those who work in retail operations should agree. The most painful thing is that the reports from different data sources do not match up. When PromptQL encounters a conflict, it doesn't make it up. Instead, it lists both sides for you to decide. Once you decide, you will use this in the future. This is very pragmatic. -
Russell_AndersonX88—Although $900 an hour for AI engineer consultation is expensive, I heard that Fortune 500 is already using it. Hasura’s approach is quite interesting, selling products and services at the same time. -
De_FiKing—The product direction has changed several times, from a data agent to a multi-person workspace, and it feels like the team is also figuring out the direction. However, the underlying technology is indeed solid, and the architecture of planning first and then executing is much more reliable than text-to-SQL. -
叶霖—If your team spends more than 5 hours per week answering repetitive data questions, PromptQL is cost-effective. But if your data stack is thin, there is no need to bother. -
Christian_Collins—To be honest, the free version is a bit limited, and you can’t check much data with 3 seats. If you want to truly evaluate its value, you still have to upgrade to Pro. But if you need multiple data sources, Pro is really cost-effective. -
孔霞超—Seeing as both McDonald's and Cisco are using it, enterprise-level security compliance shouldn't be a big problem. If SOC 2 and HIPAA are supported, the financial industry can also enter. -
MsImranMostad_88—Before, I was switching between Grafana, AWS, Snowflake, GitHub, and Linear every day. The real work was not that tiring, but the friction between the tools was tiring. PromptQL puts all of this in one place, and the flow state can finally be maintained. -
Amber.Chavez007—It can pull together data from Snowflake, Salesforce and Slack for analysis. In the past, you had to write SQL yourself to do ETL, but now you can just type it. And AI will show you its SQL and Python code, so you can rest assured.