Aegis gives the enterprise one governed path to approved OpenAI and Anthropic models through AWS. Hatchery is working directly with the Amazon Bedrock team to protect identity, company data, access, and usage at organizational scale, without storing or logging prompt and response content.
Open the live Aegis site in a new window.
At enterprise scale, personal accounts, copied keys, public model interfaces, and tool-specific settings become an unmanaged access network. Security teams cannot reliably answer who is using which model, what data boundary applies, how access is revoked, or whether sensitive work is leaving approved systems.
Source code, trade secrets, designs, customer information, contracts, strategy, and operating context can all appear in a prompt. One careless paste or unmanaged credential can create legal, security, ownership, and regulatory exposure before leadership knows it happened.
Aegis sits between familiar AI tools and approved models accessed through AWS. It binds every request to organizational identity, applies policy, controls model access, meters usage, and supports rapid revocation. Hatchery is working directly with the Amazon Bedrock team to turn that architecture into a production service people will actually use.
| Need | Direct or unmanaged access | Aegis |
|---|---|---|
| Identity | Personal accounts and static keys can outlive the employee, project, or device that received them. | Short-lived access is tied to organizational identity, making onboarding and revocation controlled operations. |
| Employee workflow | Governance often means replacing preferred tools or asking people to follow a separate manual process. | Claude Code, Codex CLI, Claude Desktop, and OpenAI-compatible clients can use one governed access path. |
| Policy | Sensitive-content rules depend on individual judgment or settings scattered across tools. | Requests follow managed policy with explicit enforcement and documented model-path boundaries. |
| Data boundary | Retention and provider behavior vary by account, product, and model path. | Content and operational metadata are deliberately separated, with model-path disclosures stated clearly for administrators. |
| Visibility | Usage, spend, and policy events are difficult to attribute across keys and services. | Identity, model, tokens, spend, timing, and policy events remain visible without retaining a transcript of employee work. |
| Scale | Every new employee, team, model, and tool adds another credential and another place policy can drift. | Access, approved models, revocation, policy, usage, and spend can be managed as one organizational system. |
Strategy, design, code, integrations, infrastructure, and operations move together so the product can launch and keep serving customers.
Developers sign in with organizational identity and receive short-lived access maintained by a local helper. There is no static model API key to copy, commit, or rotate.
Aegis connects familiar tools to approved model paths through AWS so teams can use the intelligence they need without turning every provider account into a separate security boundary.
Amazon Bedrock model access is organized into clearly described tiers so content inspection, enforcement location, output inspection, and retention expectations remain explicit.
Identity, model, token, spend, timing, and policy-event metadata support administration without turning employee AI work into a searchable transcript warehouse.
The service uses multi-AZ AWS architecture, monitored access paths, and model-routing options designed to reduce ordinary infrastructure and provider disruption without promising that outside models can never fail.
Whether the organization has 10 people or 1,000, administrators receive one place to manage access, approved models, revocation, usage, spend, and policy events without reading employee prompts.
Aegis began with a real internal requirement: our engineers needed AI tools, but Hatchery could not accept unmanaged access to code, product plans, customer context, or company knowledge.
Identity and short-lived access make onboarding, revocation, attribution, and auditability more reliable than distributing long-lived provider credentials.
Model tiers do not share identical enforcement or retention behavior. Aegis documents those differences so administrators and developers can make an informed choice.
Security teams need evidence of access, model use, spend, and policy events, not a warehouse of proprietary prompts. Aegis keeps operational metadata and excludes prompt, response, and tool-output content.
Aegis is available for organizations that need employees to use AI without creating an unmanaged data-exposure problem. It combines organizational identity, approved model paths through AWS, policy, revocation, usage and spend visibility, and multi-AZ operations in one enterprise access layer.
Aegis demonstrates what Hatchery brings beyond code: direct AWS collaboration, enterprise security thinking, product design, model integration, identity, administration, DevSecOps, and the discipline to make a safer path easier for employees to adopt.
Models are becoming commodities, but we should be able to use them without worrying about the privacy and ownership of what we build. That is why Hatchery developed Aegis.
Aegis keeps AI content moving through approved model paths while identity, policy, usage, and spend remain visible to the organization.
Organizational identity, short-lived access, policy, and approved model routes replace scattered personal accounts and static provider keys.
Administrators can see identity, model, tokens, spend, timing, and policy events while Aegis excludes prompt, response, and tool-output content.
Hatchery built Aegis first for its own safety. Our engineers needed OpenAI and Anthropic models without sending source code, product plans, customer context, and company knowledge through unmanaged accounts or static provider keys.
No. Aegis does not store or log prompt, response, or tool-output content. It records operational metadata such as user, model, tokens, spend, timing, and policy events so the organization can manage AI use without retaining employee transcripts.
Aegis supports Claude Code, Codex CLI, Claude Desktop, and OpenAI-compatible clients and SDKs. Product support can evolve, so the Aegis marketing site remains the current compatibility source.
Yes. Aegis is designed to centralize organizational identity, user access, approved models, revocation, policy, usage, and spend whether AI is used by a small technical group or 1,000 employees.
Yes. Hatchery is working directly with the Amazon Bedrock team to make Aegis practical, secure, and production-ready while providing governed access to approved OpenAI and Anthropic models through the AWS architecture.
No outside model service is infallible. Aegis uses multi-AZ AWS architecture, monitoring, approved routing, and inference-profile support where available to reduce common infrastructure and provider disruption and make failures easier to see and manage.
Current product details, security disclosures, supported tools and models, and account access are available at hatchery-aegis.com.
Clarify what to build, what to simplify, what to protect, and what must work first.
