# LaunchDarkly turns AgentControl into a runtime control plane for prompts and models LaunchDarkly has expanded AgentControl across Go and Java, Amazon Bedrock, EU data residency and conversation-level observability, making prompts, models and per-user targeting something teams can change and govern at runtime rather than redeploy with application code. AgentControl now spans more production stacks: applications can resolve different prompts and models by context, track token/cost behavior, require approvals, use Bedrock without proxying inference through LaunchDarkly, and inspect multi-step agent runs as one conversation. - Status: Active - Published: 2026-08-29T09:40:21+12:00 - Updated: 2026-08-29T09:40:21+12:00 - Categories: Artificial Intelligence, Web Development, SaaS, AI Agents, AI SaaS, Developer Tools - Tags: AgentControl, AI agents, LaunchDarkly, model routing - Canonical HTML: https://beyondthe.news/dossiers/launchdarkly-agentcontrol-runtime-prompts-models-targeting-bedrock-eu ## What changed On August 26, 2026, LaunchDarkly shipped a coordinated set of AgentControl expansions. New Go and Java AI SDKs let server applications resolve AgentControl configs at runtime, including prompts, model settings and targeting variations, while recording generation count, tokens, latency and errors. AgentControl also gained Amazon Bedrock as a provider, EU data residency for prompts/configs/targeting rules, and a Conversations view that stitches multi-step LLM and agent spans into one transcript with tool calls, token use, errors, estimated cost and evaluation scores. The core operating model is that the application still calls the model provider directly; LaunchDarkly supplies and governs the configuration rather than proxying model inference. ## Why it matters AI applications increasingly need to change prompts, models and agent instructions without coupling every experiment or emergency rollback to a software deployment. AgentControl applies LaunchDarkly's familiar targeting, approval and rollout concepts to that layer: a team can serve a model or prompt variation to selected users, change it at runtime, monitor quality and cost, and require human approval before production configuration changes. The August expansion makes that pattern usable in more server stacks and with Bedrock while adding regional-governance and run-level observability. The trade-off is maturity: LaunchDarkly's AI SDK family remains pre-1.0, and the Java SDK is explicitly not generally available, so teams should treat the control plane as evolving infrastructure rather than a frozen API contract. ## Prompts and models move out of deployment artifacts AgentControl configs contain model settings plus messages or agent instructions. Applications evaluate those configs against LaunchDarkly contexts, so different users, organizations or environments can receive different variations. Updating a prompt, model or targeting rule can take effect without rebuilding or redeploying the application. ## The August SDK expansion reaches Go and Java teams LaunchDarkly added AI SDKs for Go and Java on August 26. Both can resolve per-context AgentControl configs and report token, latency, generation and error metrics. The Go SDK is pre-1.0 and LaunchDarkly says the Java AI SDK is not yet generally available, so backward compatibility should not be assumed while these clients mature. ## Bedrock keeps inference in the customer's AWS account AgentControl can now configure Amazon Bedrock models while the application continues to call Bedrock directly using its own AWS credentials. LaunchDarkly manages prompt/model configuration and receives tracked metrics, but it does not proxy or independently invoke the model provider. That separation is useful for teams that want runtime configuration control without moving inference out of their existing AWS boundary. ## Governance comes from targeting, approvals and guarded rollouts AgentControl uses LaunchDarkly's context and segment targeting model. Teams can require approvals before config variations or targeting changes are applied and can use guarded rollouts to expose a new variation gradually while watching metrics for regressions. This turns prompt/model changes into governed releases rather than ad hoc production edits. ## Conversation views connect configuration changes to whole agent runs LaunchDarkly now groups related LLM and agent spans into conversation-level traces. A team can inspect duration, models/providers, token use, estimated cost, errors, evaluation scores and tool-call payloads across a multi-step run instead of reconstructing behavior from isolated spans. That makes it easier to compare what a rollout changed at the task level rather than only at individual model calls. ## EU residency expands where the control plane can be used AgentControl can now operate in LaunchDarkly's EU environment, keeping prompts, configs, variations and targeting rules managed in the EU and restricting model choices to providers approved for that region. Availability remains plan- and account-dependent, so teams with residency requirements need to verify both AgentControl access and their chosen provider path. ## Key details - LaunchDarkly shipped Go and Java AgentControl AI SDKs on August 26, 2026. - AgentControl configs can vary prompts, model settings and agent instructions by LaunchDarkly context or segment without an application redeploy. - The SDKs can record generation count, token usage, latency, errors and evaluation-related metrics. - Amazon Bedrock is now supported; the application calls Bedrock directly with its own AWS credentials rather than sending inference through LaunchDarkly. - AgentControl now supports EU data residency for its configuration/control data. - Conversation-level observability can show a whole multi-step agent run with tool calls, token use, estimated cost, errors and evaluation scores. - AgentControl changes can participate in LaunchDarkly approval workflows and guarded rollouts. - LaunchDarkly's AI SDK family remains pre-1.0; the Java AI SDK is explicitly not yet generally available. ## Builder takeaways - Separate prompts, model selection and agent instructions from application releases when those values need frequent experimentation or emergency rollback, but keep a reviewed fallback configuration in code for control-plane failures. - Use context targeting to run model or prompt changes on a bounded cohort before global rollout, and define task-quality plus cost metrics that determine whether the variation should expand. - Require approvals for production AgentControl changes where prompts or model routing can materially alter customer-facing behavior or tool use. - If using Bedrock, keep IAM least-privilege and verify that LaunchDarkly configuration data and model inference meet different residency/security requirements; they travel through different systems. - Treat Go/Java integrations as evolving dependencies: pin versions and run regression tests before SDK upgrades while the AI SDKs remain pre-1.0. - Review conversation traces at the completed-task level so a model that looks cheaper per call does not hide extra retries, tool calls or failed trajectories. ## What to watch - When the Go and Java AI SDKs reach stable 1.0/GA compatibility guarantees. - Additional model providers and whether provider feature differences remain portable through AgentControl configs. - Pricing and entitlement details as AgentControl adoption expands beyond account-assisted rollout. - Whether conversation observability and guarded rollouts gain stronger automatic rollback or deployment-gate integrations. - Independent evidence that runtime prompt/model targeting improves production quality or cost without introducing configuration drift. ## Uncertainties - AgentControl is an add-on whose availability depends on LaunchDarkly plan and account enablement. - The AI SDK family is pre-1.0 and the Java AI SDK is explicitly not generally available, so APIs may change. - LaunchDarkly tracks model metrics but does not invoke the model on the application's behalf; provider reliability, safety and billing remain separate operational dependencies. - EU AgentControl residency does not automatically mean every connected model provider processes inference in the same region. ## Sources - [LaunchDarkly Product Updates — AgentControl releases on August 26](https://launchdarkly.com/changelog/) — LaunchDarkly · primary changelog · 2026-08-26T00:00:00+12:00. Primary release surface for Go/Java AI SDKs, EU residency, conversation views and Amazon Bedrock support. - [AgentControl](https://launchdarkly.com/docs/home/agentcontrol) — LaunchDarkly · primary documentation. Current architecture, runtime config/targeting model, direct provider-call boundary and availability. - [AI SDKs](https://launchdarkly.com/docs/sdk/ai) — LaunchDarkly · primary documentation. Current SDK maturity, supported configuration/metric patterns and pre-1.0 status. - [Getting started with Amazon Bedrock and AgentControl configs](https://launchdarkly.com/docs/fed-docs/guides/agentcontrol/getting-started-bedrock) — LaunchDarkly · primary documentation. Bedrock integration, direct AWS inference path, config modes, tool calling and tracked metrics.