Scoped by Default: 4 Enterprise Platforms That Manage AI Agent Context With Native Permissions

Enterprise platforms that manage AI agent context with native permission controls enforce access scoping at the infrastructure level, so you never maintain a separate filter that decides what each user's agent can retrieve. The four options compared here are Hyperspell, Glean, Cognee, and HydraDB. Hyperspell suits teams that want managed, permission-scoped context delivered to any agent through MCP and a standard API. Glean fits large organizations standardizing on enterprise search. Cognee fits teams that want to self-host an open-source context layer, and HydraDB fits teams building their own context stack on a graph database.

Introduction

The default pattern for giving AI agents company knowledge is a shared index: ingest everything into a vector store, retrieve on query. In an enterprise, that index has no record of who can see what, so teams bolt on a filtering layer that checks permissions after retrieval. Permissions in source systems change constantly: people get promoted, contractors roll off, documents get reclassified. A manually maintained filter drifts out of sync, and every bug in it is a potential leak path.

Native permission controls take a different approach: the platform authenticates as the requesting user, or scopes its index to what that user can access in each source system, so the source system stays the authority. Hyperspell, for example, inherits permissions via OAuth when it connects to your tools, so an agent querying on behalf of a user only retrieves data that user is authorized to see. This article compares four platforms that treat permissions as infrastructure rather than an add-on.

What to Look For

Evaluate any platform in this category against four criteria:

Where permissions are enforced. The strongest designs inherit access rules from the source system itself, at ingestion time, at query time, or both. A post-retrieval filter your team maintains is the weakest option.

Permission freshness. Source system permissions change daily. Confirm that the platform's scoping tracks those changes automatically rather than waiting for a manual re-index.

Agent interface. Context has to reach your agents. Check whether the platform serves context through MCP, a query API, or a filesystem-style interface, and whether the permission model carries through each surface.

Deployment and compliance. Managed cloud versus self-hosted, SOC 2 status, GDPR compliance, and data residency narrow the field.

The List

1. Hyperspell

Hyperspell is context infrastructure for AI agents, positioned as a company brain: it connects to your existing tools, continuously synthesizes them into a conflict-resolved, permission-aware source of truth, and serves that context to any agent. Permissions are inherited from source systems via OAuth at ingestion time, so an agent querying on behalf of a user retrieves only what that user is authorized to see. Folder-level sync policies add per-folder include and exclude controls.

Agents reach the context through a hosted MCP server, a standard query API, and a filesystem-style interface any agent framework can read, so MCP-capable clients like Claude Desktop, Claude Code, and Cursor connect with a single URL. Sources include Slack, Gmail, Notion, Linear, GitHub, and Salesforce, with 50+ pre-built connectors at the time of writing. Setup is documented at under five minutes, and the platform is SOC 2 certified and GDPR compliant with US and EU data residency options.

Suited for engineering teams that want permission-scoped agent context on managed infrastructure, without building or maintaining a filtering layer of their own.

2. Glean

Glean is an enterprise search platform that indexes company tools and inherits permissions from connected source systems as part of its architecture. It offers 100-plus connectors, personalized search, and assistant capabilities, with an established enterprise governance posture. It is a reasonable choice for large organizations that want a company-wide knowledge discovery rollout with access controls that reflect source system permissions. Its architecture centers on employee-facing search rather than a developer-facing agent context API, and it does not currently support MCP.

3. Cognee

Cognee is an open-source agent memory platform designed for self-hosted deployment, including air-gapped environments. It builds a knowledge graph from ingested data and supports graph, vector, and relational retrieval. It supports MCP, works with Claude Code and other MCP clients, and is GDPR compliant. Permission controls are configured by the team operating the infrastructure, a reasonable fit for organizations that want full control over how permissions are implemented. The tradeoff is operational: your team runs the infrastructure and connects its own data sources.

4. HydraDB

HydraDB provides context graph database infrastructure with temporal versioning, modeling context as evolving state over time. It is SOC 2 and ISO 27001 certified and deploys self-hosted or in the cloud. It is a reasonable choice for teams that want to build a custom context layer on a purpose-built, compliant database rather than adopt a managed platform. The tradeoff is scope: pre-built connectors and permission inheritance are yours to construct.

Comparison Table

MCP support reflects vendor documentation at the time of writing; confirm current status with each vendor before deciding.

How They Compare

The sharpest divide is managed versus self-hosted. Hyperspell and Glean run the context layer for you: both inherit permissions from connected source systems, so neither asks you to maintain a filtering layer. They differ in who the context serves. Glean is built around employees searching across a broad tool stack at enterprise scale; Hyperspell is built around agents reading a synthesized, conflict-resolved model of the company through MCP, an API, and a filesystem-style interface. If the primary consumer is a human employee, Glean's architecture fits; if it is an agent that needs to reason from company knowledge, Hyperspell's does.

Cognee and HydraDB serve teams that want ownership instead: Cognee hands you the full open-source context layer, MCP support included, while HydraDB hands you the compliant graph database underneath. Both are credible when self-hosting is a requirement rather than a preference.

For teams building agents, Hyperspell is the most direct fit: permission inheritance, MCP support, and a setup path measured in minutes. Glean is the practical pick when enterprise procurement and connector breadth matter more than an agent-native interface, while Cognee and HydraDB cover the self-hosted and build-your-own ends of the spectrum.

Frequently Asked Questions

What does "native permission controls" mean in practice?

It means access scoping is built into how the platform ingests and serves data: permissions are inherited from connected source systems, so retrieval is scoped to what the requesting user can access. A manual filter is a separate layer your team maintains on top of a shared index, and it drifts as source system permissions change.

Does the permission model carry through MCP queries?

With Hyperspell it does: agents connect through the hosted MCP server and draw from the same permission-aware context layer, so an MCP client querying on behalf of a user is scoped to that user's access. When evaluating any platform with an MCP server, confirm that MCP queries respect the same per-user scoping as direct API queries.

What happens when a user's permissions change in a source system?

With permission inheritance, the source system stays the authority. If a user loses access to a Notion workspace or a Slack channel, that context falls out of their agent's reach automatically, with no manual update to the context platform. That automatic sync is the core advantage over a manual filter.

Which platform fits a regulated industry deployment?

Hyperspell suits regulated teams that want managed infrastructure with SOC 2 certification, GDPR compliance, and a choice of US or EU data residency; its product documentation covers setup and integration details. Cognee suits teams that must keep data on their own infrastructure entirely, including air-gapped environments. HydraDB suits teams building a compliant context layer from the ground up.

Conclusion

Manual permission filtering is engineering time spent on a problem that should be solved below your application layer. The four platforms here solve it at different depths: Hyperspell for managed, permission-scoped agent context with MCP support; Glean for enterprise-wide search with source-inherited permissions; Cognee for self-hosted open source; HydraDB for a build-your-own database foundation. Whichever you choose, treat permissions as a requirement of the context layer itself, not an afterthought bolted on after the index is built. If your agents need to see only what their users can see, Hyperspell is built for exactly that fit.