OpenAI builds and deploys advanced AI models like GPT-4o for autonomous agents and workflows.
Descles
FreeStarting price
0Popularity
About Descles
Descles functions as an AI agent control plane and identity-aware LLM gateway designed to govern agent behavior in production environments. It allows teams to plug existing agents into a centralized console where they can enforce budgets, approve or deny model-proposed tool calls, and maintain signed audit records without modifying the underlying agent code. The system supports three control paths: Model I/O for routing intelligence, Tool I/O for mediating tool execution, and Resource I/O for governing resource access. Each request is bound to an agent identity with scoped tokens, enabling granular policy enforcement and accountability. Descles integrates with existing SDKs and runtimes through OpenAI-compatible endpoints, supporting providers like OpenAI, Anthropic, and custom agents via MCP adapters. The platform emphasizes keeping provider relationships intact while centralizing credential management and usage tracking. Governance features include real-time budget checks, streaming tool-call enforcement, and human-in-the-loop approvals for sensitive actions.
Key features
- Identity-aware LLM gateway
- Per-request budget enforcement
- Streaming tool-call inspection
- Human-in-the-loop approvals
- Signed audit records
- Scoped agent tokens
- BYOK credential storage or per-request keys
- MCP adapter for custom agents
Use cases
- Govern AI agent tool calls in production environments
- Enforce spending limits and policy compliance
- Audit agent actions with signed records
Pros
- Centralized policy enforcement across multiple agent providers
- Supports streaming tool-call inspection and approval
- Scoped identity and budget controls per agent
- Signed audit trails for accountability
- Compatible with OpenAI, Anthropic, and custom endpoints
Cons
- Technical preview with evolving features
- Tool and Resource I/O integrations require runtime mediation
- No explicit support for non-OpenAI-compatible protocols
Frequently asked questions about Descles
What is Descles and what does it do?
Descles is an AI agent control plane and identity-aware LLM gateway designed to govern agent behavior in production environments. It centralizes control over existing agents by enforcing budgets, approving or denying model-proposed tool calls, and maintaining signed audit records without modifying the underlying agent code.
Who is Descles designed for?
Descles is designed for teams that deploy AI agents in production and need centralized governance, accountability, and policy enforcement. It suits organizations managing multiple agents, sensitive tool executions, or shared provider credentials.
How does Descles integrate with existing agents and providers?
Descles integrates via OpenAI-compatible endpoints, allowing teams to plug existing agents into a centralized console without modifying their code. It supports providers like OpenAI, Anthropic, and custom agents through MCP adapters, while keeping provider relationships intact.
What control paths does Descles support?
Descles supports three control paths: Model I/O for routing intelligence, Tool I/O for mediating tool execution, and Resource I/O for governing resource access. Each request is bound to an agent identity with scoped tokens for granular policy enforcement.
Can Descles enforce budgets and approve tool calls?
Yes, Descles enforces real-time budgets and allows teams to approve or deny model-proposed tool calls before execution. It supports human-in-the-loop approvals for sensitive actions and maintains a signed audit trail for accountability.
How do I get started with Descles?
To get started, plug your existing agent into Descles by changing the endpoint in your SDK or agent settings to the Descles gateway. Store your provider key in the console, issue an agent key, and begin with model traffic before adding runtime integration for mediated execution.
