> ## Documentation Index > Fetch the complete documentation index at: https://docs.edgeimpulse.com/llms.txt > Use this file to discover all available pages before exploring further. # Create an Edge Impulse skill > Install an Agent Skill that gives AI coding agents persistent knowledge of the Edge Impulse Studio and Ingestion APIs. AI coding agents can write scripts, automate data pipelines, trigger training jobs, and export deployments with the Edge Impulse Studio API. They can also: * Upload and relabel training data without touching the UI * Trigger and monitor training jobs from the command line * Export deployment artifacts (C++ library, Arduino, TFLite, and others) * Query project metadata, sample lists, and job status * Write bulk automation scripts for repetitive tasks like relabelling or batch uploads This tutorial walks you through installing a single [Agent Skill](https://agentskills.io/home) (`edge-impulse`) that any agent supporting the [Agent Skills specification](https://agentskills.io/specification) can load. The skill content is the same regardless of the agent; only the install path differs. Ready to see it in action? The [Motion recognition using an agent](/tutorials/end-to-end/motion-recognition-using-agent) end-to-end tutorial walks you through a complete Edge Impulse project workflow — from data upload and model training/testing, to deployment onto a device — using only natural language commands. Claude Code terminal session using the edge-impulse skill to interact with an Edge Impulse project ## Create the skill file An Agent Skill is a markdown file with required `name` and `description` fields in YAML frontmatter, followed by instructions the agent loads on demand. Copy the file below — you will save it to a different path depending on which agent you use. ````markdown SKILL.md expandable theme={"system"} --- name: edge-impulse description: Interact with an Edge Impulse project via the Studio and Ingestion APIs. Use when asked to upload data, list or relabel samples, trigger training, check job status, or export a deployment. Also use when the user asks anything about their Edge Impulse project. --- Help the user interact with their Edge Impulse project using the Studio and Ingestion APIs. ## Preparation Check for credentials in the environment before asking: - `EI_API_KEY` — Edge Impulse API key (starts with `ei_`) - `EI_PROJECT_ID` — numeric project ID If either is missing, ask the user. They can find the API key in the project Dashboard under the **Keys** tab, and the project ID under **Project Info**. ## API reference Base URL: https://studio.edgeimpulse.com/v1 Ingestion URL: https://ingestion.edgeimpulse.com OpenAPI spec: https://studio.edgeimpulse.com/openapi.yml Authentication header for all Studio API requests: ```text x-api-key: ``` ### Key endpoints **Project** ```text GET /api/{projectId} project info GET /api/ list all user projects ``` **Samples** ```text GET /api/{projectId}/raw-data list samples GET /api/{projectId}/raw-data/{sampleId} get one sample DELETE /api/{projectId}/raw-data/{sampleId} delete a sample POST /api/{projectId}/raw-data/{sampleId}/rename relabel a sample ``` Use `?category=training|testing|anomaly`, `?limit`, `?offset`, and `?labels` when listing samples. Relabel body: `{ "newLabel": "label" }` **Data upload (Ingestion API)** ```text POST /api/training/data upload training sample POST /api/testing/data upload testing sample POST /api/anomaly/data upload anomaly sample ``` Required headers: `x-api-key`, `x-label`, `x-file-name`. Accepted bodies: Edge Impulse JSON, audio/WAV, image/JPEG, CSV. Edge Impulse JSON format: ```json { "protected": { "ver": "v1", "alg": "none", "iat": 1735689600 }, "signature": "empty", "payload": { "device_name": "device", "device_type": "TYPE", "interval_ms": 10, "sensors": [{ "name": "accX", "units": "m/s2" }], "values": [[0.1], [0.2]] } } ``` **Devices** ```text GET /api/{projectId}/devices POST /api/{projectId}/device/{deviceId}/start-sampling ``` Start-sampling body: `{ "label": "...", "category": "training", "intervalMs": 10, "lengthMs": 5000, "sensor": "Inertial" }` **Impulse** ```text GET /api/{projectId}/impulse get current pipeline definition POST /api/{projectId}/impulse update pipeline ``` **Training** ```text POST /api/{projectId}/jobs/train/{versionId} start training job → returns jobId GET /api/{projectId}/jobs list jobs GET /api/{projectId}/jobs/{jobId}/status job status GET /api/{projectId}/jobs/{jobId}/stdout/download job log (plain text, not JSON) ``` Jobs are async. Poll `/jobs/{jobId}/status` until `.job.finished` is true, then check `.job.finishedSuccessful`. On failure, fetch `/jobs/{jobId}/stdout/download` and show the relevant error lines. Users can also monitor progress at `https://studio.edgeimpulse.com/studio//jobs`. **Deployment** ```text GET /api/deployment/targets POST /api/{projectId}/deploy start build — body: { "deployType": "zip"|"arduino"|"tflite"|"onnx"|... } GET /api/{projectId}/deploy get artifact download URL POST /api/{projectId}/jobs/build-ondevice-model GET /api/{projectId}/deployment/history GET /api/{projectId}/deployment/history/{deploymentVersion}/download ``` **Versions** ```text GET /api/{projectId}/versions list snapshots POST /api/{projectId}/versions create snapshot ``` **EON Tuner** ```text POST /api/{projectId}/tuner/start start AutoML run GET /api/{projectId}/tuner/runs list runs and results ``` **Testing** ```text POST /api/{projectId}/jobs/classify classify a sample GET /api/{projectId}/test-results model test results ``` **Organizations (Enterprise)** ```text GET /api/organizations GET /api/organizations/{orgId}/data/datasets POST /api/organizations/{orgId}/data/upload ``` ### Response shape Every response has a `success` field: ```json { "success": true, ... } { "success": false, "error": "message" } ``` Always check `success` before reading other fields. Raise a clear error if `success` is false. ## Python skeleton ```python import os, requests API_KEY = os.environ.get("EI_API_KEY") or input("API key: ") PROJECT_ID = int(os.environ.get("EI_PROJECT_ID") or input("Project ID: ")) STUDIO = "https://studio.edgeimpulse.com/v1/api" INGEST = "https://ingestion.edgeimpulse.com/api" s = requests.Session() s.headers.update({"x-api-key": API_KEY, "Accept": "application/json"}) def get(path, **params): r = s.get(f"{STUDIO}/{PROJECT_ID}{path}", params=params) r.raise_for_status() d = r.json() if not d.get("success"): raise RuntimeError(d.get("error", "unknown error")) return d def post(path, body=None, base=STUDIO): r = s.post(f"{base}/{PROJECT_ID}{path}" if base == STUDIO else f"{base}{path}", json=body) r.raise_for_status() d = r.json() if not d.get("success"): raise RuntimeError(d.get("error", "unknown error")) return d ``` ## Instructions 1. Understand the task from the user's request. 2. Write Python using the skeleton above, or use curl/shell commands for one-off requests. 3. Run the code and show results clearly. 4. For long-running jobs: poll `GET /api/{projectId}/jobs/{jobId}/status` until `.job.finished` is true. If `.job.finishedSuccessful` is false, fetch `GET /api/{projectId}/jobs/{jobId}/stdout/download` (plain text, not JSON) and show the relevant error lines. Always inform the user they can also monitor progress at `https://studio.edgeimpulse.com/studio//jobs`. 5. For unknown endpoints, fetch https://studio.edgeimpulse.com/openapi.yml and find the right path. ```` The `edge-impulse` skill is one large skill covering the Edge Impulse Studio and Ingestion API surface. As your usage grows, consider [splitting it into more focused skills (or companion skills)](/tutorials/topics/ai-agents/create-a-companion-skill) such as `upload-ei-data`, `label-ei-data`, `train-ei-model`, and `deploy-impulse` so each one loads less context and triggers more precisely. ## Install the skill Save the file above to one of the paths below. Skills can be installed globally (available across all projects) or per project. Refer to your agent's documentation if you use a tool not listed here. | Agent | Global (all projects) | Per project | | --------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------------- | -------------------------------------- | | [Claude Code](https://code.claude.com/docs/en/skills) | `~/.claude/skills/edge-impulse/SKILL.md` | `.claude/skills/edge-impulse/SKILL.md` | | [OpenAI Codex](https://developers.openai.com/codex/skills) | `~/.agents/skills/edge-impulse/SKILL.md` | `.agents/skills/edge-impulse/SKILL.md` | | [GitHub Copilot](https://docs.github.com/en/copilot/how-tos/copilot-on-github/customize-copilot/customize-cloud-agent/add-skills) | `~/.copilot/skills/edge-impulse/SKILL.md` or `~/.agents/skills/edge-impulse/SKILL.md` | `.github/skills/edge-impulse/SKILL.md` | | [Antigravity (previously Gemini CLI)](https://antigravity.google/docs/skills) | `~/.gemini/skills/edge-impulse/SKILL.md` | `.agents/skills/edge-impulse/SKILL.md` | ## Use the skill Ask your agent to work with your Edge Impulse project. The agent loads the skill automatically when your prompt matches, or you can invoke it explicitly with `/edge-impulse`: /edge-impulse list all training samples labeled "unknown" and delete them /edge-impulse upload every .wav file in ./data as testing samples with label "noise" /edge-impulse start a training job and wait for it to finish, then show accuracy /edge-impulse export a C++ library deployment and save it to ./build To see a selection of example prompts, take a look at our [Prompt library](/tutorials/topics/ai-agents/prompt-library). Claude Code terminal session using the edge-impulse skill to interact with an Edge Impulse project ### Set your credentials Set environment variables so your agent never needs to ask for credentials: ```bash theme={"system"} export EI_API_KEY="ei_..." export EI_PROJECT_ID="12345" ``` Find your [API key](/studio/projects/dashboard/api-keys) under **Dashboard → Keys** and your [project ID](/studio/projects/dashboard#6-project-info) under **Project Info** in Edge Impulse Studio. Claude Code terminal running the /edge-impulse skill ## Optimize the skill for your workflow If you find yourself making the same API calls repeatedly — listing samples, checking job status, fetching project info — ask the agent to add those patterns directly to the skill file so it can handle them faster next time: Read \~/.claude/skills/edge-impulse/SKILL.md and add a section at the bottom with a ready-made Python snippet for listing all training samples grouped by label, since I use this at the start of every session. Look at my last few Edge Impulse prompts and update \~/.claude/skills/edge-impulse/SKILL.md with any patterns I use frequently that aren't already covered. The skill file is just markdown — anything you add becomes part of the agent's persistent context for every future session. [Keep `SKILL.md` under 500 lines.](https://code.claude.com/docs/en/skills#add-supporting-files) Move detailed reference material — long endpoint tables, full request/response examples, language-specific snippets — into separate files in the skill directory (such as `reference.md` or `examples.md`) and link to them from `SKILL.md`. Supporting files load only when the agent decides it needs them, so the main skill stays focused and cheap to load. ## Tips * **Ask for error handling.** The Edge Impulse API returns `{ "success": false, "error": "..." }` on failure. Ask the agent to check this field in every response. * **Reference the OpenAPI spec for edge cases.** For endpoints not covered by the skill, direct the agent to `https://studio.edgeimpulse.com/openapi.yml` for the full schema. ## Next steps * Follow the [end-to-end Motion recognition tutorial](/tutorials/end-to-end/motion-recognition-using-agent) to take a project from data collection through deployment using only natural language commands * [Create a companion skill](/tutorials/topics/ai-agents/create-a-companion-skill) for your target hardware so the agent can write application code alongside the API calls * Browse the [prompt library](/tutorials/topics/ai-agents/prompt-library) for ready-to-use prompts covering every stage of an Edge Impulse project * Browse [Studio API endpoints](/apis/studio) for the full endpoint reference