OpenBB Open Data Platform Review 2026: Open-Sourced, AGPL, and What It Means for Quants
OpenBB Open Data Platform Review 2026
On August 25, 2026, OpenBB founder and CEO Didier Lopes published a post titled “OpenBB belongs to everyone” announcing that the entire OpenBB product suite would be open-sourced under a permissive open source license. The company is winding down. For quantitative traders, data scientists, and ML engineers who have followed OpenBB’s trajectory from terminal to platform, this is the most consequential event in the project’s history — and the one that demands the most careful reading.
Sources and scope: This review is based on desk research from the official OpenBB documentation site, the OpenBB GitHub repository, the PyPI package index, and the published OpenBB pricing page, all accessed on 2026-09-23. We did not install, run, or hands-on test any component of the Open Data Platform. No benchmarks, no trial deployments, no code execution. Everything stated here about behaviour, pricing, and licensing is drawn from the sources cited inline.
What Changed on August 25, 2026
OpenBB announced it is open-sourcing its full product suite under a permissive licence and winding down the company, with founder Didier Lopes stating the team could not find product-market fit within the time available. The licence is a promise in a blog post, not a change in the repository, which is the distinction that matters most for anyone planning to build on this code.
The announcement named four products being open-sourced: OpenBB Workspace, OpenBB Copilot, the OpenBB Excel Add-in, and the Open Data Platform. In the post, Lopes wrote: “we couldn’t find the product-market fit needed to build a sustainable business around this vision within the time we had.” The company confirmed it is winding down operations. No specific permissive licence was named in the announcement — no Apache, no MIT, no BSD. The GitHub repository’s LICENSE file still reads “Copyright (c) 2021-2025 OpenBB Inc. All files in this repository are licensed under the GNU Affero General Public License v3.0.” The permissive licence is announced, not landed.
This distinction is critical. The GitHub repository, described as “Open Data Platform for analysts, quants and AI agents,” carries 73,411 stars and 7,598 forks as of this writing, with the default branch named develop and a last push on 2026-09-23. The PyPI package openbb sits at version 4.7.2, released 2026-05-26, and still carries the classifier “GNU Affero General Public License v3.” The sibling openbb-core package is at version 1.6.13, released 2026-06-17. The legal transition from AGPL to permissive has been promised but has not yet been executed in the code that ships.
OpenBB announcement · GitHub repository · PyPI openbb · PyPI openbb-core
What the Open Data Platform Is
The Open Data Platform is an infrastructure layer designed to let teams connect to data sources once and consume that data across multiple surfaces: a Python SDK for quants, a Workspace and Excel interface for analysts, MCP servers for AI agents, and REST APIs for other applications.
According to the official ODP documentation, ODP is “the open-source toolset that helps data engineers to integrate proprietary, licensed, and public data sources into downstream applications like AI copilots and research dashboards.” It is built around a “connect once, consume everywhere” philosophy. The platform exposes four surfaces: Python for quants, OpenBB Workspace and Excel for analysts, MCP servers for AI agents, and REST APIs for other applications. Three components deliver these surfaces: ODP Desktop, a standalone desktop app for managing Python environments and backend servers; ODP Python, the PyPI packages providing SDKs, REST APIs, and MCP servers; and ODP CLI, a command-line interface wrapping the installed packages.
For a quant team, the architecture matters because it means a single data integration can feed a Jupyter notebook, an Excel dashboard, an AI copilot, and a custom web application simultaneously. The documentation positions ODP as the plumbing layer beneath the OpenBB Copilot and Workspace products, not as a standalone analytics tool. Understanding this positioning is essential before evaluating whether the platform fits a given workflow.
The Python Surface
The Python surface is the primary interface for quantitative users, built around a single import statement and a path-based menu system that routes calls to different data providers, with provider-dependent parameters and third-party extensions available through an “all” extra. For a quant, that means the same research code keeps working when you swap a free provider for a licensed one, at the cost of coping with provider-specific interval and field coverage.
The quickstart documentation shows the canonical entry point: from openbb import obb. From there, the platform uses a path-based menu — obb.equity.price.historical, obb.equity.fundamental, obb.equity.screener, obb.equity.estimates, obb.equity.ownership, obb.equity.shorts, obb.equity.darkpool, obb.equity.discovery, obb.equity.compare, obb.equity.calendar, obb.equity.profile, obb.equity.market_snapshots, and obb.equity.search. The obb.equity.price.historical endpoint alone supports providers including alpha_vantage, cboe, fmp, intrinio, polygon, tiingo, tmx, tradier, and yfinance. Critically, interval choices differ per provider, so the documentation advises consulting the docstring for each provider extension. Third-party provider extensions are installable via the “all” extra.
Here is the primary Python snippet from the quickstart:
from openbb import obb
# Fetch historical equity prices
data = obb.equity.price.historical(
symbol="AAPL",
provider="yfinance"
)
# Convert to a pandas DataFrame
df = data.to_dataframe()
print(df.head())The path-based design means that a quant can switch providers by changing a single parameter — from provider="yfinance" to provider="polygon" — without rewriting the call structure. This is the core value proposition for data engineers who need to abstract provider-specific logic behind a uniform interface. The provider-dependent parameter variations are a known friction point: not every provider supports every interval, every field, or every endpoint, and the docstring is the authoritative source for what is available.
Python quickstart · Provider extensions
REST and MCP Surfaces for AI Agents and Applications
The REST API surface lets any HTTP-capable application consume ODP data, while the MCP server surface allows AI agents to call ODP tools directly, making the platform usable beyond Python-native workflows. Both surfaces are generated from the same provider extensions, so an agent and a notebook query the same normalised schema rather than two separately maintained integrations.
The documentation describes the REST API as the surface for “other applications” — web apps, dashboards, or services that need market data without a Python runtime. The MCP server surface is designed for AI agents, allowing copilots and agentic workflows to invoke ODP data tools as native function calls. This dual-surface approach means a data engineering team can expose the same underlying data integrations to both a Python-based quant workflow and an AI agent workflow without building separate pipelines.
The REST API reference documents the surface as a FastAPI application generated from the same Python packages, so a local deployment looks like this:
# The REST surface is a FastAPI app generated by the Python packages.
uvicorn openbb_core.api.rest_api:app # serves http://127.0.0.1:8000
curl -X GET "http://127.0.0.1:8000/api/v1/equity/price/historical?symbol=AAPL&provider=yfinance" \
-H "accept: application/json"This pattern — a versioned endpoint, a symbol parameter, and a provider parameter — is generated directly from the Python signature, which is why the REST path mirrors obb.equity.price.historical. For teams building AI agent workflows, the MCP server means the agent can call obb.equity.price.historical directly through the model context protocol rather than routing through a custom REST wrapper. The practical implication is that ODP positions itself as the data layer beneath both human-facing analytics tools and agentic systems, which is a meaningful architectural choice for teams building multi-surface data products.
Deployment and Pricing at Time of Writing
At the time of writing, OpenBB offers a free Community tier, a Lite tier at $2,400 per year (promoted at $1,200), a custom-priced Pro tier, and a Snowflake-hosted option at $500 per seat per year, with Workspace Lite available for self-hosted small teams.
The pricing page lists four tiers. Community is free, individual licence, cloud-hosted by OpenBB. Lite is $2,400 per year at list price, shown at $1,200 per year under a 50%-off promotion that ran through August 31; it is a team licence, self-hosted on-prem or in a VPC, aimed at small teams under ten people. Pro is custom pricing, team licence, self-hosted at scale. Snowflake is $500 per seat per year, Snowflake-hosted. The pricing table includes OpenBB Copilot, Workspace MCP server, Excel add-in, RBAC, MFA/IdP, audit logs, and an App Marketplace as feature rows.
Workspace Lite, launched on 2026-07-21, is a self-hosted package for deployments of roughly one to twenty people, requiring at least one data source and Docker. The documentation explicitly states it is not a Kubernetes-based enterprise deployment. This makes Lite the relevant option for small quant teams that want self-hosted control without enterprise orchestration overhead.
A critical caveat: the pricing page predates the August 25 shutdown announcement. The post-shutdown commercial direction — whether pricing will change, whether cloud-hosted Community will continue, whether self-hosted Lite will remain the primary offering — is unstated. Any team evaluating pricing today should treat the published numbers as the last known commercial terms, not as a forward-looking commitment.
OpenBB pricing · Workspace Lite announcement · Workspace Lite docs
Licensing: The Permissive Promise vs the AGPL Reality
The announced permissive licence has not yet landed in the code; the repository and PyPI package remain under AGPL-3.0, which imposes copyleft obligations on any hosted or modified deployment. If you modify the platform and offer it to users over a network, the licence requires you to publish those modifications under the same terms.
The gap between the announcement and the legal reality is the single most important licensing fact for any team evaluating ODP. The August 25 post promises a permissive open source license. The GitHub LICENSE file still reads AGPL-3.0. The PyPI classifier still reads “GNU Affero General Public License v3.” The platform moved to AGPL in a 2024 license change, and that change has not been reversed in the code as of this writing.
What this means in practice depends on who you are. AGPL-3.0 is a strong copyleft licence: if a fund or vendor modifies ODP and offers it as a hosted service, the licence requires releasing those modifications under AGPL-3.0. A permissive licence — MIT, Apache-2.0, BSD — would impose no such obligation. For a fund embedding ODP in an internal research platform with no external hosting, AGPL’s network-use clause is less of a concern. For a vendor building a commercial product on top of ODP and hosting it for clients, AGPL-3.0 is a significant legal consideration that a permissive licence would eliminate.
Until the licence actually changes in the repository and on PyPI, any legal review must proceed on the assumption that AGPL-3.0 remains in force. The announcement is a statement of intent, not a executed license change.
OpenBB announcement · GitHub repository · PyPI openbb · 2024 license change
Who Should and Should Not Adopt It Now
Teams that need a self-hosted, multi-surface data integration layer and can accept AGPL-3.0 obligations should evaluate ODP now; teams that need a commercially supported, permissively licensed platform should wait until the licence transition is executed. The decision turns on two variables you can check yourself: the LICENSE file in the repository, and whether your deployment is internal or hosted for third parties.
Adopt now if: You are a small quant team or data engineering group that needs a uniform Python interface across multiple data providers, values the “connect once, consume everywhere” architecture, is comfortable with AGPL-3.0 for internal or self-hosted use, and can tolerate the maintenance risk of a project whose original company is winding down. The 73,000-plus GitHub stars suggest a substantial community that may continue development, but community maintenance is not the same as commercial support.
Wait if: You are a vendor or fund that needs to embed ODP in a commercially hosted product and cannot accept AGPL-3.0 copyleft obligations. Wait until the permissive licence is actually executed in the repository and reflected on PyPI. If your legal team requires a permissive licence for any hosted deployment, the current AGPL-3.0 state is a blocker regardless of the announcement’s intent.
The bus-factor risk is real. A company winding down means the core engineering team is dispersing, the product roadmap is frozen, and commercial support is ending. The GitHub repository shows active commits as of 2026-09-23, which suggests community or maintenance activity, but the long-term trajectory is unknowable. Any adoption decision should include a contingency plan for maintaining the platform internally if community development stalls.
ODP Scorecard
Scores reflect desk research against the sources cited above, weighted for a quant team evaluating ODP for production research infrastructure.
| Dimension | Score | Notes |
|---|---|---|
| Data-provider coverage | 9/10 | Fifteen-plus equity providers behind one call, with per-provider interval and field gaps to absorb |
| Python API design | 9/10 | Single import, path-based menu, provider swap by one parameter |
| Licensing clarity | 4/10 | Permissive licence announced, AGPL-3.0 still in the repository and on PyPI |
| Commercial support outlook | 3/10 | Company winding down; community maintenance is not a support contract |
| Multi-surface reuse | 8/10 | Python, REST, MCP and Excel generated from the same provider extensions |
| Documentation quality | 8/10 | Quickstart and provider pages are current; post-shutdown direction is unstated |
| Overall Score | 7/10 | Adopt for internal, self-hosted use if you accept AGPL-3.0 and bus-factor risk |
FAQ
Is the Open Data Platform free to use?
The Community tier is free, individual licence, cloud-hosted by OpenBB. However, the company is winding down, so the longevity of the cloud-hosted Community tier is uncertain. Self-hosted options (Lite at $1,200–$2,400 per year, Pro at custom pricing) require Docker and at least one data source. The free tier’s future depends on whether community infrastructure replaces the company’s cloud hosting.
What Python version does ODP require?
The PyPI package openbb requires Python >=3.10 and <4. The package version at time of writing is 4.7.2, released 2026-05-26. The openbb-core sibling package is at version 1.6.13, released 2026-06-17.
Can I use ODP with AI agents?
Yes. The documentation describes an MCP server surface specifically designed for AI agents, allowing copilots and agentic workflows to invoke ODP data tools as native function calls. The REST API surface also enables any HTTP-capable application to consume ODP data. Both surfaces are part of the ODP Python component, delivered through PyPI packages.
← Back to all posts
