Pyra designs, deploys, and manages Claude agents for GTM, sales, marketing, and operations teams. According to Anthropic's documentation, the Claude Agent SDK puts the agent loop that powers Claude Code into your own application, and Claude Managed Agents runs long-running, asynchronous agents on Anthropic-hosted infrastructure.We turn that into scoped, permissioned workflows your team can approve — starting small and expanding only what proves out.
Want an agent wired into the GTM stack over MCP that owns a repeatable outcome — research, prep, and drafts — with permission rules you control.
Need account research, meeting prep, and drafted outreach produced in the background and left for review — nothing sends without sign-off.
Want long-running, multi-step workflows that survive pauses and resume cleanly, with data boundaries designed per workflow and audit trails throughout.
Every workflow below is grounded in documented platform capabilities and configured with permission rules and human review. Pyra sells the implementation and management service.
The following facts are drawn from Anthropic's documentation, verified August 18, 2026. Where a claim is made by the vendor, we attribute it as such.
Per Anthropic's documentation, the Agent SDK provides the same tools, agent loop, and context management that power Claude Code, programmable in Python and TypeScript — with built-in tools, hooks, subagents, sessions that resume or fork, and skills and memory.
Anthropic's documentation describes permission controls over which tools run automatically and which need approval, and MCP for connecting external tools and data. In June 2026, Anthropic added enterprise-managed authorization so admins can centrally authorize MCP connectors for their organization through their identity provider.
Per Anthropic, Claude Managed Agents is a hosted REST API where Anthropic runs the agent and its sandbox — built for long-running and asynchronous work, with streaming events, mid-execution steering and interrupts, and a choice of cloud sandbox or self-hosted sandbox on your own infrastructure.
Anthropic's documentation states Managed Agents is stateful by design — conversation history, sandbox state, and outputs are stored server-side — and is not currently eligible for Zero Data Retention. Pyra's deployment work includes designing data boundaries around this: what enters a session, when to use a self-hosted sandbox, and when to build on the Agent SDK in your own infrastructure instead.
We pick one workflow with a clear result, choose the right surface — Agent SDK in your infrastructure or Managed Agents — and start with read-and-prepare work.
We connect your tools over MCP, set permission rules for which tools run automatically and which need approval, and design data boundaries around what each session can reach.
We monitor, tune, and expand — widening permissions or adding workflows only once results hold up in review, with audit trails throughout.
No. Pyra is an independent implementation partner. We are not affiliated with, sponsored by, or endorsed by Anthropic. We design, deploy, and manage Claude agent workflows for our clients. Claude, the Claude Agent SDK, and Claude Managed Agents are products of Anthropic.
Both, depending on the workflow. Per Anthropic's documentation, the Agent SDK is a Python and TypeScript library that runs the same agent loop, tools, and context management that power Claude Code — inside your own process and infrastructure. Claude Managed Agents is a separate hosted product: Anthropic runs the agent and the sandbox behind a REST API, built for long-running and asynchronous work, with the option of a self-hosted sandbox on your own infrastructure. We match the surface to your data requirements and the shape of the workflow.
Per Anthropic's documentation, Claude Managed Agents is stateful by design: sessions are long-running, resume after pauses, and conversation history, sandbox state, and outputs are stored server-side. Anthropic's documentation also states that Managed Agents is not currently eligible for Zero Data Retention. Because of this, our deployment work includes designing data boundaries: choosing what enters a session, using a self-hosted sandbox where requirements call for it, or building on the Agent SDK in your own infrastructure instead.
Start Free to scope one repeatable workflow with a clear result. From there we build the agent, connect your tools over MCP, set permission rules for which tools run automatically and which need approval, and pilot before expanding — with monitoring and audit trails throughout.
Start Free to scope a single read-and-prepare workflow with a clear result. Then Pyra builds the agent, connects your tools over MCP, sets permission rules and data boundaries, and manages the deployment as it expands. Independent implementation partner — not affiliated with Anthropic.