--- title: "Integrations" source: https://docs.autohand.ai/integrations/ --- # Connect Autohand to your stack Choose a work surface, agent protocol, or model provider. Each guide shows the credentials, configuration, supported workflows, and constraints you need before connecting Autohand. Quick answer ## What are Autohand integrations? Autohand integrations connect the coding agent to developer tools, automation clients, and model runtimes. They let teams bring repository context into a workflow, control agents programmatically, or select where inference runs. Status Developer platforms, ACP and RPC, cloud model APIs, and local model runtimes In development Linear, Sentry, and Datadog native integrations Requires Provider-specific for hosted services; local runtimes can work without a cloud API key Configure with Environment variables, Autohand config, or the integration's setup flow Directory ## Choose an integration Start with the system you need Autohand to reach. ### Work surfaces Review code, manage work, and respond where your team already collaborates. [SlackBring codebase questions, incident response, and engineering assistance into Slack. Workspace install](/integrations/slack.html) [GitHubAutomate pull-request review and CI workflows with GitHub Actions. GitHub token](/integrations/github.html) [BitbucketConnect repositories, pull requests, and Bitbucket Pipelines. App password or OAuth](/integrations/bitbucket.html) [JiraLink engineering work to Jira issues and automate issue updates. API token or OAuth](/integrations/jira.html) [LinearPlanned native issue and project workflow integration. In development](/integrations/linear.html) [SentryPlanned error analysis and debugging workflow integration. In development](/integrations/sentry.html) [DatadogPlanned monitoring and incident correlation integration. In development](/integrations/datadog.html) ### Agent protocols Embed or control an Autohand session from another application. [ACP ModeMaintain a session-oriented, bidirectional connection for editors and agent clients. ACP client + token](/integrations/acp-mode.html) [RPC ModeControl Autohand with newline-delimited JSON over stdin and stdout. Local process](/integrations/rpc-mode.html) ### Cloud model providers Use a hosted model API and keep the provider choice explicit in Autohand. [Compare all model providersSee configuration patterns, tradeoffs, and local alternatives in one guide. Provider overview](/integrations/ai-model-providers.html) [OpenRouter Route to supported model APIs through one provider connection. OPENROUTER\_API\_KEY](/integrations/openrouter.html) [OpenAI Connect Autohand directly to an OpenAI API account. OPENAI\_API\_KEY](/integrations/openai.html) [AWS Bedrock Use Bedrock models with AWS identity, regional, and network controls. AWS credentials or API key](/integrations/aws-bedrock.html) [Azure AI Foundry Use a deployed Azure model with Azure authentication and controls. Azure endpoint + credentials](/integrations/azure-foundry.html) [GCP Vertex AI Run supported models through a Google Cloud project and region. Google Cloud credentials](/integrations/gcp-vertex.html) [DeepSeek Use DeepSeek models through its OpenAI-compatible API. DEEPSEEK\_API\_KEY](/integrations/deepseek.html) [xAI Grok Connect Autohand to Grok models with the xAI provider. XAI\_API\_KEY](/integrations/xai.html) [Z.ai Use GLM-family models through an OpenAI-compatible endpoint. ZAI\_API\_KEY](/integrations/zai.html) [Sakana AI Use supported Fugu orchestration models from Autohand. SAKANA\_API\_KEY](/integrations/sakana.html) [Cerebras Connect to the Cerebras OpenAI-compatible inference API. CEREBRAS\_API\_KEY](/integrations/cerebras.html) [NVIDIA Use NVIDIA-hosted models through its compatible endpoint. NVIDIA\_API\_KEY](/integrations/nvidia.html) [LLMGateway Connect through a gateway that fronts multiple model providers. LLMGATEWAY\_API\_KEY](/integrations/llmgateway.html) ### Local model runtimes Keep inference on hardware you control, with no hosted model API key required. [Ollama Run a local Ollama server and select an installed model. Ollama host](/integrations/ollama.html) [MLX Run supported models locally on Apple Silicon. Apple Silicon](/integrations/mlx.html) [llama.cpp Serve a GGUF model from local or self-managed hardware. llama.cpp server](/integrations/llama-cpp.html) Setup model ## How integrations work 1. 01 **Choose the connection** Pick the developer tool, protocol, or model runtime that owns the workflow. 2. 02 **Provide only the required access** Create the documented token, app installation, cloud identity, or local endpoint. Keep secrets out of source control. 3. 03 **Configure and verify** Apply the guide's environment variables or config, then run its smallest verification step before expanding permissions. Before you connect ## Availability and limitations Provider model catalogs, identifiers, quotas, and regional availability can change independently of Autohand. Confirm those details in the provider console before production use. Linear, Sentry, and Datadog are documented as in development and do not yet have a public production setup. Common questions ## Integration FAQ ### Which Autohand integration should I choose? Choose a work-surface integration for team workflows, ACP or RPC when another application must control an agent, and a model-provider integration when you need to select where inference runs. ### Do all Autohand integrations require an API key? No. Hosted services usually require provider credentials, while local runtimes such as Ollama, MLX, and llama.cpp can run without a cloud model API key. Each guide lists its exact requirement. ### Where should integration credentials be stored? Use the environment variable, secret store, or Autohand configuration method named in the integration guide. Do not commit API keys, app passwords, or cloud credentials to a repository. ### Are every listed integration and provider available? No. Linear, Sentry, and Datadog are currently marked in development. Hosted provider availability also depends on the provider account, model catalog, region, and quota.