Introduction - AC2Documentation IndexFetch the complete documentation index at: /llms.txtUse this file to discover all available pages before exploring further.Skip to main contentAC2 home pageSearch...⌘KAsk Assistant⌘IDashboardDashboardSearch...NavigationGetting startedIntroductionGetting startedIntroductionQuickstartPlatformDatasetsEvaluationTrainingServingBring your own harnessDeployment modelsFAQCLIOverviewProjectsRunsAnalysisResourcesServingSDKClient overviewRuntimeRunsAnalysisResourcesServingTracingMCPOverviewToolsOn this pageOnboard with your coding agentPackagesChoose a workflowGetting startedIntroductionCopy pageCopy pageEnd-to-end infrastructure for training and inferencing open weight models.Copy pageCopy pageApplied Compute Agent Cloud (AC2) is a researcher cockpit for end-to-end model development, inference and continual learning.
​Onboard with your coding agent
The fastest way to get started is to let your coding agent set AC2 up for you. Paste this prompt into Claude Code, Cursor, Codex, or any coding agent:
Fetch https://platform.appliedcompute.com/agents.md (e.g. curl -fsSL https://platform.appliedcompute.com/agents.md) and follow its instructions to set up AC2 for me.

The agent installs the CLI, logs you in, wires up the AC2 MCP server, and walks you through your first project. Prefer to run the steps yourself? Follow the Quickstart.
​Packages

ac2.sdk — platform client for datasets, workloads, sessions, traces, and secrets.
ac2.runtime — agent framework for agents, environments, orchestrators, tasks, graders, and episodes.
ac2.tracing — OpenTelemetry-based observability SDK. Add @traced to any function to capture timing, inputs, outputs, and nested spans. Exports to ClickHouse and Langfuse.

QuickstartInstall the SDK and train a model.RuntimeDefine agents, environments, orchestrators, and eval tasks.SDKManage datasets, deployments, runs, traces, and secrets from Python.MCP serverAnalyze eval and train traces from your editor.
​Choose a workflow
Evaluate an agentRun evals, grade traces, and compare behavior.Train a modelChoose GRPO, supervised fine-tuning, or on-policy self-distillation.Route inferenceDeploy a model and route traffic through Policies and Endpoints.FAQFind answers for failed runs, datasets, package installation, and cleanup.QuickstartgithubPowered byThis documentation is built and hosted on Mintlify, a developer documentation platformAssistantResponses are generated using AI and may contain mistakes.

The Applied Compute Agent Cloud (AC2) functions as an end-to-end infrastructure designed for the development, inference, and continual learning of open weight models, positioning itself as a researcher cockpit. Access to the complete documentation index is available at /llms.txt, which serves as a starting point for exploring the platform's various sections, including the dashboard, navigation, CLI, MCP Overview, and tools.

The recommended entry point for users to begin interacting with AC2 is through onboarding with a coding agent. This agent automates the setup process by installing the command-line interface (CLI), handling user login, wiring up the AC2 MCP server, and guiding the user through an initial project. Alternatively, users can follow the Quickstart guide manually to achieve the same result.

AC2 is structured around several specialized packages that define its functionalities. The ac2.sdk serves as the platform client, providing access to crucial resources such as datasets, workloads, sessions, traces, and secrets. The ac2.runtime encompasses the agent framework, enabling the management of agents, environments, orchestrators, tasks, graders, and episodes. Furthermore, the ac2.tracing package is an observability SDK built on OpenTelemetry principles, allowing users to capture detailed information like timing, inputs, outputs, and nested spans for functions, exporting this data to ClickHouse and Langfuse.

The platform supports various workflows that developers can utilize for different phases of model development and deployment. These include evaluating an agent by running evaluations, grading traces, and comparing behavioral results. For model development, AC2 supports training methodologies such as GRPO, supervised fine-tuning, or on-policy self-distillation. Inference capabilities are managed through routing traffic via defined Policies and Endpoints.

The overall architecture is supported by centralized resources, including a dedicated section for running analysis, tracking resources, and tracing across the system. Frequently Asked Questions address common issues related to package installation, dataset management, failure recovery for runs, and general cleanup procedures. The documentation is hosted on Mintlify, emphasizing its role as comprehensive developer documentation.