MindsDB
From an open-source “AI inside the database” project to a federated data-and-context layer for AI agents — and, in 2026, the company behind the MindsHub agent workspace.
Company Overview
| Field | Verified / best-supported answer |
|---|---|
| Company | MindsDB, Inc. / MindsDB |
| Industry | AI infrastructure, data infrastructure, developer tools, open-source software |
| Founded | 2017 according to YC and historical company profiles; first version released in late 2018 according to the current company story. |
| Headquarters | Historically Berkeley / San Francisco Bay Area. Current company materials describe a remote-first, globally distributed team; the current contact page lists San Jose, California as the HQ mailing address. |
| Company type | Private |
| Official/current web presence | mindshub.ai; legacy mindsdb.com URLs redirect to MindsHub. |
| Countries of operation | Worldwide / globally distributed. An exact country count is Not Publicly Available. |
| Mission | Democratize AI by making AI easier to use while keeping infrastructure open and inspectable. |
| Tagline / positioning | Historically: “Connect, Unify, Respond…” and “AI Data Vault”. Current product message: a home/workspace for open-source AI agents. |
Why MindsDB matters
MindsDB attacks an unusually persistent AI bottleneck: organizations often have useful data in many databases, SaaS systems and documents, while AI models live in separate infrastructure. The company’s core thesis is that the expensive part is frequently not the model itself but connecting models to trustworthy, permissioned, live context.
Its strategic evolution is therefore important. The company began by putting predictive ML close to databases, expanded into LLM and retrieval workflows, and now positions its query/context engine as infrastructure for agents that need to reach distributed enterprise data.
Founders & Team
Jorge Torres — Co-founder & CEO
- Education: Australian National University; public profiles also reference postgraduate/technical study and UC Berkeley research involvement. Exact degree chronology is not consistently published.
- Career: Worked across data-intensive startups and engineering roles; public profiles cite Skillshare, Real Life Analytics and CareJourney.
- Expertise: machine learning, distributed/data systems, ML automation, explainability, product strategy.
- Role: Co-founder and CEO; current company materials list him in leadership.
- Notable: Visiting scholar/researcher work at UC Berkeley on ML automation and explainability.
Adam Carrigan — Co-founder & COO
- Education: Australian National University; University of Cambridge MPhil; public sources describe research focused on NLP and equity-price prediction.
- Career: Research roles, Deloitte UK management consulting, entrepreneur-in-residence work, and co-founder of Real Life Analytics.
- Expertise: operations, strategy, finance, marketing, NLP, commercialization and open-source community building.
- Role: Co-founder and COO in company profiles.
- Previous company: Real Life Analytics, a computer-vision/digital-signage venture founded with Torres.
Founding Story
The founding story combines a technical insight with an unusually strong open-source philosophy. Torres and Carrigan met at the Australian National University and later worked together on a previous computer-vision business. They became interested in the gap between the growing power of machine learning and the limited number of people able to build and deploy it.
The initial 2017 product thesis was straightforward: instead of forcing every organization to assemble a bespoke data-science stack, put machine-learning capabilities next to the database and expose them through a familiar interface. The original announcement described automatic construction and training of deep-learning predictive models, with an extended SQL syntax for querying them.
The company’s current history says the first version was released in late 2018 in Berkeley. It was inspired by the “Minds” in Iain M. Banks’ Culture novels — intelligent systems that collaborate with people rather than replace them. That philosophical framing became a useful product principle: bring intelligence to the systems people already use.
Company founded / initial product announced. Core thesis: add predictive intelligence to existing databases.
First version released as an open-source project in Berkeley; the “Minds” concept from Iain M. Banks becomes part of the identity.
Y Combinator Winter 2020; $3M financing announced; reported developer/download traction supports the open-source strategy.
Expansion of database partnerships and seed funding; Snowflake, SingleStore and DataStax integrations highlighted.
$16.5M Series A led by Benchmark; additional $25M financing led by Mayfield; $5M NVIDIA/NVentures investment later in the year.
Shift toward LLMs, knowledge bases, retrieval, AI agents and broader data integrations.
v26.0 and v26.1 strengthen the federated query/context engine, Knowledge Bases, agents, integrations and security.
MindsDB becomes the parent identity and MindsHub becomes the agent-facing product surface. Existing workloads continue without migration.
Company Timeline
| Year | Milestone | Significance |
|---|---|---|
| 2017 | Founding and first public product concept | Established the “AI closer to data” thesis. |
| 2018 | Open-source first version | Created a developer-led distribution engine. |
| 2020 | Y Combinator W20; $3M financing | Shifted from project to venture-backed company. |
| 2021 | Database ecosystem partnerships and $7.6M cumulative seed funding reported | Expanded reach through existing data platforms. |
| 2022 | Recognition as Gartner Cool Vendor; continued ML platform expansion | External validation of the data/AI infrastructure thesis. |
| Feb 2023 | $16.5M Series A, Benchmark | Commercialization and enterprise growth phase; Forbes reported $56M valuation. |
| Jun 2023 | $25M additional financing, Mayfield-led | Company said total funds raised reached $50M. |
| Aug 2023 | $5M NVIDIA/NVentures investment | Company said total seed funding reached $46.5M; strategic AI-infrastructure alignment. |
| 2024 | Agents, LLMs and community expansion | Positioning moved beyond classic predictive ML. |
| Feb 2026 | MindsDB v26.0 | Federated data/context engine for AI applications and agents. |
| Apr 2026 | v26.1 and product updates | Broader integrations, reliability, Knowledge Base fixes and shareable outputs. |
| May 2026 | MindsHub rebrand | Agent workspace becomes the primary product surface; MindsDB remains the parent identity. |
Products & Services
1. MindsDB Query / Data Engine
Purpose: Federated access to many data sources through a SQL-compatible interface, with AI-specific constructs. Users: developers, data engineers, AI engineers and analytics teams. Features: connectors, SQL querying, Knowledge Bases, agents, jobs/triggers, model integrations. Technology: Python, SQL parser/planner, connectors, vector/semantic retrieval and external AI engines. Pricing: self-hosted/open-source options plus commercial services; current public hosted pricing is centered on MindsHub rather than a legacy MindsDB seat plan.
2. Knowledge Bases
Purpose: combine structured data with unstructured/vectorized content for retrieval and grounded AI. Users: enterprise AI builders. Features: semantic search, metadata filtering and retrieval across documents/data. Strength: reduces the need to build a separate retrieval pipeline for every source. Risk: retrieval quality depends on data quality, indexing, permissions and model behavior.
3. AI Models / Predictors
Historically, MindsDB’s AI Tables let users create predictors and query predictions as though they were database tables. Supported use cases included classification, regression, time-series forecasting and NLP. The modern platform also connects to external LLMs and model frameworks rather than requiring one proprietary model.
4. MindsHub
Current agent-facing product surface introduced in 2026. It is a workspace for delegating projects to open-source AI agents, publishing results and connecting agents to data/tools. The underlying MindsDB engine remains part of the stack rather than disappearing.
5. Anton / Hermes
Open-source agent harnesses used by MindsHub. Anton is the default current harness; Hermes is an alternative. The strategic goal is model/agent portability rather than a single locked runtime.
Representative competitors
Technology Stack & Technical Architecture
Architecture in plain English
MindsDB can be understood as a translation and orchestration layer between data sources and AI models/agents. The user connects a data source, defines or invokes an AI capability, and then queries or automates the result through a familiar interface. The current architecture is explicitly described as Connect → Unify → Respond.
Federated access to databases, SaaS systems, files and other sources. Public documentation covers systems such as PostgreSQL, MySQL, SQLite, Google Cloud SQL, Cloud Spanner, IBM Db2, MongoDB and many others.
Combine structured tables with vectorized/unstructured context in Knowledge Bases. This is important for RAG-style applications and semantic search.
Run models or agents against the unified context and return structured answers, predictions, search results or actions.
Technology areas
| Area | What is verified |
|---|---|
| Programming | Public GitHub repositories show Python as the dominant implementation language; supporting Shell, Dockerfile, HCL and other tooling is present. |
| AI / ML | Classical ML, deep learning, time-series models, NLP, LLMs, embeddings, semantic search and agent workflows. |
| LLMs | Integrations have included OpenAI, Anthropic, Hugging Face, Google Gemini, LangChain and other model providers/frameworks. |
| Computer Vision | Computer vision was part of the founders’ earlier Real Life Analytics work, but it is not a core current MindsDB product category. A dedicated MindsDB CV product is Not Publicly Available. |
| Speech AI | Not Publicly Available as a dedicated core MindsDB product. |
| Reinforcement Learning | Not Publicly Available as a core MindsDB product. |
| Cloud | Hosted cloud plus self-hosted deployment. Current materials emphasize on-prem, VPC, serverless and hybrid deployment patterns. |
| APIs | HTTP and MySQL APIs have been core interfaces; PostgreSQL and SDK-based access are also documented. A JavaScript SDK was introduced for web applications. |
| Data | Enterprise databases, warehouses, SaaS applications, files and unstructured documents, depending on connector. |
| Security | Credentials vault concepts, security policy, responsible disclosure process and ongoing vulnerability patching. SOC 2 Type II is listed on current public materials. |
| Privacy | The architectural value proposition is that data can remain in existing systems rather than being copied into a new AI silo; however, each connector/model deployment must still be governed for permissions, retention and provider-specific data policies. |
Business Model
MindsDB has historically used a classic open-core/enterprise infrastructure model: open-source software creates adoption, developer trust and distribution, while commercial cloud, support, enterprise deployments and managed infrastructure provide monetization.
Revenue mechanisms
- Hosted/managed AI infrastructure and cloud services.
- Enterprise support and commercial deployments.
- Private/on-prem/VPC and custom solutions.
- Usage of managed model/agent infrastructure in the current MindsHub product.
Distribution
- Open-source GitHub distribution.
- PyPI/Docker installation.
- Database and AI framework integrations.
- Developer tutorials, events and hackathons.
- Enterprise partnerships and direct sales.
Product Evolution Timeline
AI directly adjacent to databases; open-source first.
Y Combinator W20 and $3M financing; early adoption validates the developer-led distribution strategy.
Database partnerships, $7.6M cumulative funding reported and broader predictive use cases.
Gartner Cool Vendor recognition; platform positioned as applied ML infrastructure.
Benchmark $16.5M Series A; Mayfield $25M financing; NVIDIA $5M investment; LLM integrations and agent-oriented experiments accelerate.
Knowledge Bases, semantic search, LLM connectors, agent workflows and broader enterprise data integrations become central.
v26.0/v26.1 strengthen the federated context engine; May 2026 rebrand makes open-source agents the primary product story.
Growth Strategy
Growth flywheel
- Open-source adoption: lower the barrier to trying the technology.
- Integrations: make the platform useful wherever developers already store data.
- Community: tutorials, Slack/Discord, GitHub contributions, events and hackathons generate feedback and distribution.
- Enterprise conversion: organizations that need governance, SLAs, security and managed infrastructure become potential commercial customers.
- Expansion: new AI models and agent frameworks increase the value of the same data-access layer.
This is closer to developer-led infrastructure growth than consumer viral growth. The company’s open-source community is therefore a strategic asset: it creates product feedback, integration breadth and credibility before an enterprise sale.
Funding History
| Date | Round / event | Amount | Investors / evidence |
|---|---|---|---|
| 2018–2020 | Early / accelerator capital | Included in later cumulative figures | Berkeley SkyDeck, Y Combinator and early investors appear in company/press histories. |
| Apr 2020 | Financing | $3M | Led by OpenOcean; other early investors included Berkeley SkyDeck and private investors. VentureBeat reported the round and early adoption. |
| Nov 2021 | Seed financing / extension | $3.75M announced; $7.6M cumulative reported | Walden Catalyst Ventures joined existing backers including Y Combinator, OpenOcean, Speedinvest and Berkeley SkyDeck. |
| Feb 2023 | Series A | $16.5M | Led by Benchmark; Chetan Puttagunta joined the board. Forbes reported a $56M valuation. |
| Jun 2023 | Additional financing | $25M | Led by Mayfield with TQ Ventures and existing investors; company said total funds reached $50M. |
| Aug 2023 | Strategic investment | $5M | NVentures/NVIDIA-led; company said total seed funding reached $46.5M. |
Competitive Landscape
| Dimension | MindsDB / MindsHub | Databricks | Snowflake | DataRobot / H2O.ai |
|---|---|---|---|---|
| Primary wedge | Federated AI/data context + agents | Data/AI lakehouse platform | Cloud data platform with AI | ML/AI development and operationalization |
| Open source | Core projects are open source | Selective open source; commercial platform | Primarily commercial | H2O has major open-source roots; DataRobot is commercial |
| Data federation | Core design principle | Strong, but often centered on its lakehouse/data plane | Strong ecosystem, often centered on Snowflake | Usually requires data integration pipelines |
| LLM/agent layer | Core current direction | Strong | Strong | Strong |
| SQL-centric developer UX | Core differentiator | SQL is important | SQL is core | Less central |
| Self-hosting | Strong | Enterprise deployment options | More cloud-centric | Enterprise deployment |
| Pricing | Open source + commercial; current MindsHub usage pricing | Enterprise consumption model | Consumption/contract model | Enterprise contracts |
| Strategic risk | Open-source monetization + intense agent competition | Platform complexity / competition | Platform concentration | Model commoditization / platform competition |
Market-position conclusion: MindsDB is not best understood as a direct replacement for every enterprise data platform. Its strongest niche is the connective tissue between heterogeneous enterprise data and AI/agent systems — particularly where teams value SQL, open source, federation and self-hosting.
SWOT Analysis
Strengths
- Open-source distribution and strong developer ecosystem.
- Broad integration strategy.
- Neutrality across data stores and model providers.
- SQL-compatible interface lowers adoption friction.
- Long history connecting AI to real enterprise data.
- Current shift toward agents aligns with market demand.
Weaknesses
- Smaller scale than major cloud/data-platform competitors.
- Open-source usage does not automatically convert to revenue.
- Product story changed significantly over time, creating positioning complexity.
- Security incidents can be especially consequential for a data-access layer.
- Limited public financial transparency as a private company.
Opportunities
- Enterprise agent adoption.
- MCP and agent interoperability.
- RAG and semantic search over distributed data.
- On-prem and sovereign AI deployments.
- AI governance and permission-aware context layers.
- Model commoditization increases value of orchestration and data access.
Threats
- Databricks, Snowflake, AWS, Google and Microsoft adding similar capabilities.
- Agent frameworks becoming mature enough to make custom integration easier.
- Foundation-model vendors moving down the stack.
- Security vulnerabilities or supply-chain concerns.
- Open-source competitors with stronger funding or community momentum.
AI & Innovation
MindsDB's innovation is primarily systems-level rather than a proprietary frontier model. The company has repeatedly tried to make AI capabilities accessible through the data layer.
- In-database ML: predictive models exposed through database-like abstractions.
- AutoML: automated model-building workflows designed for developers and data users.
- Explainability: an explicit theme in early product and research work.
- LLM integration: connectors for model providers and frameworks such as OpenAI, Hugging Face, Gemini and LangChain.
- Time-series: integrations with Nixtla/StatsForecast and other forecasting engines.
- Knowledge Bases: semantic retrieval across structured and unstructured information.
- Agents: data-connected autonomous workflows and, in 2026, the MindsHub agent workspace.
- Open source: public repositories include the main query engine, SQL parser, examples, SDKs and current agent projects.
Patents: A verified, comprehensive patent portfolio was Not Publicly Available in the authoritative sources reviewed. The company’s defensibility appears to rely more on open-source adoption, integrations, developer experience and accumulated infrastructure know-how than on a publicly documented patent moat.
Partnerships & Ecosystem
Representative ecosystem relationships
- Database / data: Snowflake, SingleStore, DataStax and a broad set of SQL/NoSQL/cloud connectors.
- AI/model ecosystem: OpenAI, Anthropic, Google Gemini, Hugging Face, LangChain, Nixtla and other model/framework providers.
- Capital partners: Benchmark, Mayfield, Y Combinator, NVIDIA/NVentures, OpenOcean, Walden Catalyst Ventures and others.
- Community: GitHub contributors, Discord/Slack communities, hackathon participants and open-source maintainers.
Specific government contracts or government collaborations are Not Publicly Available in the authoritative sources reviewed. The founders’ earlier work with CareJourney involved large-scale healthcare data systems and Aneesh Chopra, but that should not be confused with a current government partnership.
Global Presence
North America, Europe, Middle East, South America and Asia Pacific.
Current repository/product materials describe 200+ integrations/data sources.
No authoritative fixed country count is published.
The company began in Berkeley and built a globally distributed developer community. Current materials describe a remote-first organization with physical bases in the US and UK. Localization is primarily technical: connectors, APIs, SQL compatibility, documentation and support across different enterprise environments rather than a country-by-country consumer strategy.
Marketing Strategy
- Developer-first branding: technical clarity, open-source identity and strong emphasis on “AI + data”.
- Content marketing: tutorials, technical explainers, integration guides and use cases are a major acquisition channel.
- SEO: large integration/documentation surface creates many search-entry points for specific AI/data problems.
- Community: GitHub, Slack/Discord, contributor programs and hackathons create feedback and awareness.
- Events: the company has hosted developer gatherings; a 2024 hackathon attracted 250 hackers.
- Partnership marketing: database vendors and AI-framework integrations expand distribution.
- Enterprise sales: open-source adoption can identify teams that later need security, governance, support and managed deployments.
This strategy is structurally different from a consumer AI company: credibility among developers and data teams is the primary brand asset.
Company Culture
Culture signals
- Open-source by design.
- Distributed/remote-first operating model.
- International contributor base.
- Emphasis on developer autonomy and experimentation.
- Community participation as part of the product strategy.
Benefits / hiring
Public third-party profiles have historically listed benefits such as flexible/remote work, wellbeing support, insurance and retirement matching. Exact current benefits and the full hiring process are Not Publicly Available and may change.
Current company materials emphasize hiring across regions rather than requiring a single office location.
Awards & Recognition
Forbes
Recognized on the AI 50 list in 2021 according to company and YC profiles.
Gartner
Recognized as a Cool Vendor for Data and AI in 2022 according to company/YC materials.
Fast Company
Current company history lists Fast Company recognition among its selected external validations.
Challenges, Controversies & Security
| Issue | What happened | Status / lesson |
|---|---|---|
| CVE-2022-23522 | Unsafe archive extraction could allow files to be written outside an intended directory. | Fixed in later versions; demonstrates risk of remote model/file workflows. |
| CVE-2023-49795 and related advisories | SSRF and file-write issues were identified by GitHub Security Lab. | Security hardening became a necessary part of the platform lifecycle. |
| CVE-2024-24759 | DNS rebinding could bypass SSRF protections. | Patched in 23.12.4.2. |
| CVE-2025-68472 | Unauthenticated path traversal in file upload could expose filesystem data. | Fixed in 25.11.1 according to NVD. |
| CVE-2026-27483 | Authenticated path traversal in /api/files could enable remote code execution. | Fixed in 25.9.1.1 according to NVD; illustrates why secure-by-default file handling matters. |
Other controversies: No major verified legal controversy or public ethical scandal was found in the authoritative sources reviewed. The principal public risk is technical: a platform that connects AI agents to enterprise systems has a high security blast radius if permissions, file handling or network access are misconfigured.
Future Roadmap & Outlook
Officially observable direction: MindsDB is moving toward becoming the data/context and agent infrastructure underneath open-source AI work. The May 2026 rebrand is the clearest strategic signal: the company says the product surface changed because the old “DB” name no longer described work involving documents, vectors, agent runtimes and tool-using workflows.
Likely strategic priorities — evidence-based
- Expand agent-to-enterprise-data connectivity.
- Make MCP a first-class interoperability layer.
- Improve Knowledge Bases and retrieval quality.
- Support more models and agent harnesses.
- Strengthen on-prem, VPC and hybrid deployments.
- Improve security and observability for production agents.
Analyst forecast — clearly labeled
- Data permissions and governance will become a larger differentiator than raw model capability.
- Model routing and portability may become a core commercial feature.
- Open-source distribution will remain important for enterprise trust and developer adoption.
- The biggest challenge will be converting infrastructure breadth into repeatable enterprise revenue.
Key Metrics
| Metric | Value | Confidence / note |
|---|---|---|
| Founded | 2017 | YC/company profiles; first version late 2018 per current company story. |
| Founders | Jorge Torres, Adam Carrigan | Confirmed by YC and company materials. |
| Current company identity | MindsDB; MindsHub is the current agent product surface | Confirmed by May 2026 rebrand announcement. |
| Ownership | Private | Confirmed by startup/company databases; no public listing. |
| Funding | Over $55M | Current company/YC materials; exact cap table not public. |
| Valuation | Not currently public | $56M reported at the 2023 Series A; $160M was reported in a 2023 profile, neither is a current valuation. |
| Employees | Not Publicly Available; third-party sources generally place it in the 11–50 range | Do not treat directory headcount as audited. |
| Open-source deployments | 500K+ claimed in May 2026 | Company-reported. |
| GitHub stars | ~39K | Observed on public repository in August 2026; dynamic metric. |
| Data integrations | 200+ | Current repository/company materials. |
| Revenue | Not publicly disclosed | CB Insights lists $1M for 2023; third-party historical estimate only. |
| Countries served | Worldwide | Exact country count Not Publicly Available. |
| HQ / mailing | San Jose, California (current mailing address); historical Berkeley/San Francisco roots | Current contact page. |
Lessons for Entrepreneurs
1. Solve an integration bottleneck
Models are increasingly commoditized. The durable pain may be connecting them to governed, live enterprise context.
2. Meet users where they already work
SQL was not invented for AI, but it was already understood by millions of developers and data professionals. That reduced adoption friction.
3. Open source can be a distribution engine
Open source creates trust, experimentation and community. But founders must design the conversion path to paid value early.
4. Build neutrality into infrastructure
Supporting many data systems and models can be more strategically durable than betting everything on one vendor.
5. Security becomes product strategy
When software can reach databases, files and agent tools, security failures are not edge cases. They can define enterprise adoption.
6. Reposition without abandoning the core
The 2026 rebrand shows a useful pattern: preserve the underlying engine while changing the product surface when the old category name becomes constraining.
7. Community events create product intelligence
Hackathons and contributor programs are not only marketing; they reveal emerging workflows and integration priorities.
8. Avoid proprietary-model dependency
A model-neutral orchestration layer can remain valuable as model providers change rapidly.
Discussion Questions & References
20 discussion questions
- Why did MindsDB choose the database as the initial AI integration point?
- Was SQL the right abstraction for democratizing ML? Why or why not?
- How does open-source distribution change the economics of enterprise AI infrastructure?
- What would be the strongest moat for MindsDB if foundation models become commodities?
- Should MindsDB focus on developers, data teams, or business users?
- What are the advantages and disadvantages of federated querying compared with ETL into a central warehouse?
- How should an AI data layer enforce permissions across multiple systems?
- What security architecture is required before an agent can safely write to production systems?
- Is MCP likely to become a durable standard or another transitional protocol?
- How should MindsDB measure the ROI of its open-source community?
- What is the best enterprise pricing model for an open-source AI infrastructure company?
- Should the company remain model-neutral if one model provider becomes clearly dominant?
- What can the 2026 MindsHub rebrand teach founders about category positioning?
- Which competitors could most easily replicate MindsDB's functionality?
- Does self-hosting increase enterprise trust enough to justify the operational complexity?
- How should MindsDB communicate security vulnerabilities without damaging developer trust?
- What is the best path from 500K+ open-source deployments to durable recurring revenue?
- Would an acquisition by a cloud/data platform create more value than remaining independent?
- What metrics should an investor use to evaluate MindsDB if revenue is not public?
- What should founders learn from MindsDB's evolution from predictive ML to agents?