Quick answer
Choose the generally available Microsoft Discovery platform for organizational R&D programs that need governed enterprise data, coordinated agents, scientific tools, repeatable workflows, and an operating model. Choose the Microsoft Discovery app preview for local, earlier-stage exploration using a GitHub Copilot account.
The shared brand does not make the surfaces interchangeable.
| Decision | Discovery platform | Discovery app |
|---|---|---|
| Lifecycle | Generally available | Preview |
| Primary user | Organization and R&D program | Researcher, student, lab, small team |
| Starting access | Enterprise Microsoft relationship and current service setup | GitHub download and eligible GitHub Copilot account |
| Primary context | Governed institutional knowledge, agents, modeling, simulation, analysis, validation | Local working environment, literature, hypotheses, reasoning, experiments |
| Operating burden | Enterprise identity, data, tool, compute, governance, support, and change management | Local security, dependency, repository, compute, data, Copilot, and preview changes |
The Microsoft Discovery product guide provides the full entity and source record.
Select by project maturity
Use the app when a bounded individual or small-team investigation needs a practical starting point and local context. Record the app version, repository revision, dependencies, Copilot account, environment, data sources, prompts, outputs, and validation.
Move toward the enterprise platform when the work requires shared institutional data, persistent governance, coordinated specialist agents, managed scientific tools, cross-team review, formal access, or production support. Do not postpone identity, data, source, safety, or reproducibility controls simply because the app is local.
Local does not mean no data leaves
“Local app” describes the working experience, not a complete data-flow guarantee. Document what stays on the machine, what is sent to models or services, what external sources and packages are called, which logs or telemetry exist, and how GitHub Copilot terms apply.
Minimize sensitive and proprietary inputs, use approved machines and accounts, and verify retention, deletion, region, contracts, repository security, dependency provenance, and outbound network access.
Migration checklist
Before moving a local investigation into the GA platform, preserve:
- question, hypothesis, scope, and decision criteria;
- source identifiers, retrieval dates, licenses, and access rights;
- data snapshots, transformations, code, packages, and environment;
- model, agent, tool, connector, and configuration versions;
- outputs, reviewer notes, failed branches, and limitations;
- validation data, experiments, statistics, and domain approvals.
Re-run critical results in the target environment. A copied transcript or artifact is not evidence that tools, data, versions, permissions, and compute behaved identically.
The local app access guide covers the preview setup. The Discovery Engine workflow guide covers enterprise iteration design.
Frequently asked questions
Is Microsoft Discovery generally available?
The enterprise Microsoft Discovery platform became generally available on June 2, 2026. The separate Microsoft Discovery desktop app was announced in preview on the same date, so the two surfaces have different lifecycle claims.
Who should use the Discovery app preview?
Microsoft positions the app for researchers, students, academic labs, and scientific teams starting local literature exploration, hypothesis generation, reasoning, and iterative experimentation without a full enterprise deployment.
Can app work move to the enterprise platform?
Microsoft says locally developed work can move into the Discovery platform as projects mature. Treat that as a supported direction, then verify live export, import, identity, source, environment, connector, permission, and reproducibility details before relying on migration.
Bottom line
The GA platform is an organizational R&D service; the app is a local preview entry point. Choose lifecycle and operating model first, then preserve enough evidence to reproduce any work that moves between them.
Official source
Source check: August 2, 2026. Verify current lifecycle, access, repository, Copilot, systems, pricing, and migration documentation.