# AI use cases for Experiential (experientiallabs/experiential) Repo: [experientiallabs/experiential](https://github.com/experientiallabs/experiential) | Use cases: 3 | Page: https://dedupedai.com/use-cases/experientiallabs-experiential/ Experiential is an open source gateway and router for agent workflows. It puts hosted, BYOK and local models behind one OpenAI-compatible API, lets you control which users and agents can use which models and how much they can spend, and turns production traffic into a custom router or model optimized for quality, speed and cost. This file holds every use case of this repo. Each one starts with a "## " heading and ends with its URL. - Run one local gateway in front of all your models: https://dedupedai.com/use-cases/experientiallabs-experiential/run-one-openai-compatible-gateway-for-all-your-models/ - Repoint Claude Code, Cursor and Codex at one gateway: https://dedupedai.com/use-cases/experientiallabs-experiential/point-your-coding-agents-at-the-hosted-gateway/ - Turn your agent traces into a cheaper router: https://dedupedai.com/use-cases/experientiallabs-experiential/turn-agent-traces-into-a-cheaper-router/ ## How to run one local gateway in front of all your models with Experiential Repo: [experientiallabs/experiential](https://github.com/experientiallabs/experiential) | Difficulty: Beginner | Time: 15 minutes | Basis: From the README Start the gateway with exp, get a one-time key and send a normal chat request to a public alias. **Who it is for:** A developer who switches between providers and wants a single OpenAI-compatible endpoint with a budget on it. **The problem:** Every tool wants its own provider key and URL, and nothing stops one runaway agent from burning through money. **The outcome:** A local OpenAI-compatible gateway that answers requests for an alias you chose, with a command budget attached. ### Steps 1. Run `pip install experiential` 2. Run `exp` to start a local OpenAI-compatible gateway; the setup wizard picks provider, model and reasoning effort, then shows defaults for the public alias, identity and the `$50.00` command budget before printing a one-time key 3. Choose a public alias such as `opus-5` and capture the issued key 4. Set the key with `export EXP_GATEWAY_KEY=...`, then send the curl request from the README to `http://127.0.0.1:8000/v1/chat/completions` with your alias as the model and the key in the Authorization header ### Prompt to paste into your agent ```text Install experientiallabs/experiential with `pip install experiential`, start the local gateway with `exp`, and walk me through the setup wizard. Pick the provider and model I choose, create the public alias , then send the curl test request from the README and show me the response. ``` ### Caveat The README shows the local gateway on 127.0.0.1:8000 only; check the docs before exposing it to other machines. Risk, credentials: The gateway holds your provider keys and issues a key of its own, so keep both out of git and out of shared shells. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/experientiallabs-experiential/run-one-openai-compatible-gateway-for-all-your-models/ ## How to repoint Claude Code, Cursor and Codex at one gateway with Experiential Repo: [experientiallabs/experiential](https://github.com/experientiallabs/experiential) | Difficulty: Intermediate | Time: 30 minutes | Basis: From the README Hand your coding agent the SETUP.md prompts so it connects your provider keys and repoints your tools. **Who it is for:** A team lead whose people use several coding agents and who wants usage and spend in one place. **The problem:** Each agent talks to a different provider with a different key, so there is no single view of what is being spent. **The outcome:** Your existing coding agents calling the gateway, with your own provider keys connected for pass-through. ### Steps 1. Open the hosted platform at platform.experientiallabs.ai, which serves the same OpenAI-compatible (and Anthropic Messages) API at `https://api.experientiallabs.ai/v1` 2. Open SETUP.md in the repo and copy one of its four prompts into your coding agent (Claude Code, Cursor, Codex and similar) 3. Pick Connect your inference provider keys (BYOK) to attach your own OpenAI, Anthropic, Gemini, Azure, Bedrock, Fireworks or OpenRouter keys for free pass-through 4. Or pick Full onboarding to create an account from your email, connect your keys, import your spend, then repoint every coding agent or Conductor at the gateway 5. Make a first `/v1` call with the OpenAI or Anthropic SDK using an `xpl_` key to confirm it works ### Prompt to paste into your agent ```text Read SETUP.md from experientiallabs/experiential and run its Connect your inference provider keys (BYOK) prompt for me. Ask before creating any account or pasting any key, then repoint at the hosted gateway and make one test call. ``` ### Caveat The hosted platform is a separate service from the open source gateway, with its own account. Risk, credentials: Setup connects real provider keys and creates an account on a hosted service, so review each step your agent proposes. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/experientiallabs-experiential/point-your-coding-agents-at-the-hosted-gateway/ ## How to turn your agent traces into a cheaper router with Experiential Repo: [experientiallabs/experiential](https://github.com/experientiallabs/experiential) | Difficulty: Advanced | Time: 2 hours | Basis: From the README Build a simulation from your traces, optimize a router against it and fine-tune a model you own. **Who it is for:** An engineer running a support or coding agent in production who wants lower cost without losing quality. **The problem:** Picking models by gut feel leaves money on the table, and nobody has data on which cheaper model would have been good enough. **The outcome:** A project router fitted to your own traffic, and optionally a fine-tuned open source model. ### Steps 1. Collect OpenTelemetry traces from your current agent, or try the public terminal-tasks OTLP dataset on Hugging Face, which the README downloads with a curl command as traces.otel.jsonl 2. Run `exp build support-agent` and follow the prompts for providers, models, budget and your trace file 3. Load the fitted router from Python with `exp.load_router("my-project")` as an official OpenAI client backed by its own private gateway 4. After collecting traces from your router, run `exp optimize model support-agent` to fine-tune an open source model you own using Tinker ### Prompt to paste into your agent ```text Using experientiallabs/experiential, download the public terminal-tasks trace dataset from the README, run `exp build` on it for a project called , and tell me which models and budget it chose and why. Do not run `exp optimize model` until I approve. ``` ### Caveat The README does not state expected savings, so measure cost and quality on your own traces before you switch traffic. Disclaimer: Third-party repos we don't own or vouch for. AI-generated use cases can be wrong: review everything and run at your own risk. Full text: https://dedupedai.com/disclaimer/ URL: https://dedupedai.com/use-cases/experientiallabs-experiential/turn-agent-traces-into-a-cheaper-router/