> For the complete documentation index, see [llms.txt](https://docs.multitudes.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.multitudes.com/integrations/open-telemetry.md). # Open Telemetry (OTel) ## What is OTel? OTel is an open-source, vendor-neutral standard for generating, collecting, and exporting telemetry data such as metrics, logs, and traces. Multitudes uses OpenTelemetry to receive your AI tooling usage data. For more on OpenTelemetry, see the [official documentation](https://opentelemetry.io/docs/what-is-opentelemetry/?utm_source=chatgpt.com). ## How it works ### What metrics we can ingest and process We can ingest and process OTel data for: * Claude Code [Team](https://claude.com/pricing/team) and [Individual](https://claude.com/pricing) plans * Codex If you're on the Claude Code [API](https://claude.com/pricing#api) or [Enterprise](https://claude.com/pricing/enterprise) plan, head to our docs here: [Claude Code](https://docs.multitudes.com/integrations/claude-code). {% hint style="info" %} If you'd like to bring in Open Telemetry data from other AI tools besides Claude Code or Codex, let us know at . {% endhint %} ### Integration options We can ingest OTel metrics in two different ways: #### 1. Multitudes-hosted collector Claude Code and Codex users send metrics directly to the Multitudes Open Telemetry collector running in Multitudes infrastructure. This avoids having to host a collector in your own infrastructure. To do this, follow steps here: [Multitudes-hosted collector](#multitudes-hosted-collector). Or jump to specific instructions for: * [Claude Code](#for-claude-code) * [Codex](#for-codex) #### 2. Self-hosted collector You either run your own collector or the Multitudes Open Telemetry collector in your infrastructure. The collector ingests metrics from Claude Code and Codex users and forwards them to Multitudes. To do this, follow the steps below: [Self-hosted collector](#self-hosted-collector) ## Multitudes-hosted collector When using the Multitudes-hosted collector, each user sends their telemetry data directly to the Multitudes endpoint. For this, each user needs a configuration file that specifies the endpoint to use and an `Authorization` header. Apply the following configuration to each user's machine: #### For Claude Code: 1. Open the `~/.claude/settings.json` . 2. Add the following lines, using an API key that has been generated (following the instructions below): ```json "env": { "CLAUDE_CODE_ENABLE_TELEMETRY": "1", "OTEL_METRICS_EXPORTER": "otlp", "OTEL_LOGS_EXPORTER": "otlp", "OTEL_EXPORTER_OTLP_PROTOCOL": "http/protobuf", "OTEL_EXPORTER_OTLP_ENDPOINT": "https://integrations.multitudes.co/otel", "OTEL_METRIC_EXPORT_INTERVAL": "10000", "OTEL_EXPORTER_OTLP_HEADERS": "Authorization=Bearer " }, ``` {% hint style="warning" %} We’re starting new work to measure AI effectiveness. If you’d like to help us out and be beta testers for these upcoming features, include the following in your `~/.claude/settings.json`: **Tool usage logging** ``` "OTEL_LOG_TOOL_DETAILS": "1" ``` This turns on detailed logging of tool usage in telemetry traces, which includes Bash commands, MCP server and tool names, skill names, prompt length, and the input arguments passed to each tool. This redacts the contents of prompts unless you have specifically enabled that (see next section). **Bonus: Prompt logging** ``` "OTEL_LOG_USER_PROMPTS": "1" ``` If you like, you can also turn this on to send us the contents of prompts that your organization is using. We do not receive the contents of any attached files unless they are explicitly included in the prompt text. More about these environment variables in the [Claude Code docs](https://code.claude.com/docs/en/monitoring-usage). Note that enabling both of these settings is optional and will not change anything in the Multitudes app now. But enabling this data will give us real user data to test new features on and you’ll gain early access to these new features as we build them. {% endhint %} {% hint style="info" %} The `Authorization` header can be an Organization API key if you are provisioning this file via MDM and want to use the same key across multiple users. If you would prefer that each user has their own key, they can create a Personal API key and update their `~/.claude/settings.json` file. {% endhint %} {% hint style="info" %} **AWS Bedrock:** when using Bedrock the `user.email` attribute must be configured manually via the `OTEL_RESOURCE_ATTRIBUTES` environment variable. This is because Bedrock authentication does not automatically provide user email information. Each user will need to add their work email address (or whichever email is associated with their user in Multitudes): ```json "env": { ... "OTEL_RESOURCE_ATTRIBUTES": "user.email=developer@company.com" } ``` {% endhint %} #### For Codex: 1. Open the `~/.codex/config.toml`. 2. Add the following lines: ```toml [otel] exporter = { otlp-http = { endpoint = "https://integrations.multitudes.co/otel/v1/logs", protocol = "binary", headers = { "Authorization" = "Bearer " } } } log_user_prompt = false ``` 3. Each user should [generate a personal API key](/configuration-and-setup/api-keys.md) 4. Replace `` with your actual API key. Restart any sessions you might have in Claude. {% hint style="info" %} **Authenticated via API Key:** If logging in or authenticating Codex using an API key (regardless if it's the desktop app or the CLI), each team member will also need to add the following to their shell profile (e.g. `~/.zshrc` or `~/.bashrc`), since API key auth doesn't carry account identity the way ChatGPT login does: ```bash export OTEL_RESOURCE_ATTRIBUTES="user.email=developer@company.com" ``` {% endhint %} ## Self-hosted collector When using the self-hosted collector, each user sends their telemetry data to a collector run in your infrastructure, from which aggregated metrics are sent to Multitudes. ### Running the collector 1. Set up the Multitudes OTel Collector. Detailed instructions can be found [here](https://github.com/MultitudesCo/otel-collector). 2. [Follow the instructions to generate an Organization API key](/configuration-and-setup/api-keys.md) and run the collector using this key. ### Sending from an existing collector 1. If you have an existing collector, update your config to add an exporter that sends to the Multitudes ingestion endpoint. The main sections to configure are in `exporters` , `processors`, and `service`. 2. Add a new `otlphttp/multitudes` exporter ```yaml exporters: otlphttp/multitudes: endpoint: https://integrations.multitudes.co/otel encoding: json headers: Authorization: "Bearer ${env:MULTITUDES_API_KEY}" timeout: 30s retry_on_failure: enabled: true initial_interval: 5s max_interval: 30s max_elapsed_time: 300s ``` 3. Add the new `otlphttp/multitudes` to your existing metrics and logs pipelines: ```yaml service: pipelines: metrics: receivers: [...your existing receivers] processors: [...your existing processors] exporters: [...your existing exporters, otlphttp/multitudes] logs: receivers: [...your existing receivers] processors: [...your existing processors] exporters: [...your existing exporters, otlphttp/multitudes] ``` 4. Follow the instructions below to generate an Organization API key and run the collector using this key. ### Sending data to the collector Apply the following configuration to each user's machine: ```json { "env": { "CLAUDE_CODE_ENABLE_TELEMETRY": "1", "OTEL_METRICS_EXPORTER": "otlp", "OTEL_LOGS_EXPORTER": "otlp", "OTEL_EXPORTER_OTLP_PROTOCOL": "http/protobuf", "OTEL_EXPORTER_OTLP_ENDPOINT": "", "OTEL_METRIC_EXPORT_INTERVAL": "10000" } } ``` {% hint style="info" %} **AWS Bedrock:** when using Bedrock the `user.email` attribute must be configured manually via the `OTEL_RESOURCE_ATTRIBUTES` environment variable. This is because Bedrock authentication does not automatically provide user email information. Each user will need to add their work email address (or whichever email is associated with their user in Multitudes): ```json "env": { ... "OTEL_RESOURCE_ATTRIBUTES": "user.email=developer@company.com" } ``` {% endhint %} ## Checking data ingestion status Once you are sending data to Multitudes, either directly or via the self-hosted option, you can check the ingestion status in the Multitudes app, Open the [OpenTelemetry integration modal](https://app.multitudes.co/team-settings/integrations?pill=true\&integration=otel) to check whether your data is coming through.
Once data is being received successfully, your Claude usage metrics will appear in Multitudes after our batch processing pipeline runs. Our pipeline runs daily, so in most cases you should expect to see metrics populate in the Multitudes App within up to one day of sending data.