Forward Deployed Engineer, requirement 7 of 12
API and integration design
I publish the API contract first, gate it with keys and scopes, then build integrations against it.
The requirementThe role needs strong knowledge of APIs, enterprise integrations and automation, including designing the APIs and integrations.
How I deliver it at a customer
- Map the systemsI list the systems that must talk, how each one authenticates, the events it emits and who owns it. That map decides what becomes a published endpoint and what stays an internal call.
- Publish the contractI publish the contract before anyone writes an integration. The Luca API serves an OpenAPI 3.1 document at GET /v1/schema, and one per app at /v1/apps/{slug}/schema. The document is generated from the code that serves the routes, so the contract cannot drift. Every endpoint, scope and response shape sits there, so client teams read the contract, not my code.
- Gate every callI gate every route on the bearer key, the scope the endpoint declares, and a per-key bucket of 120 requests per minute. A key without the declared scope gets a 403 that names the missing scope, and no partial result. Over the limit the caller gets a 429 with Retry-After, and another account's app returns the same 404 as a missing one. Every reply uses one envelope with success, errors, result and a request id, so any call can be traced later.
- Expose tools to modelsI register platform operations as model tools with declared parameters and required fields, through the agent tool bridge. The bridge runs nothing that is not registered, so a model cannot reach an undeclared operation. Destructive operations never run on the model's word; they go to a human approval step first.
- Wire the automationI put event-triggered CRM workflows behind documented HTTP trigger routes, and the management endpoints behind role checks. When a role check fails the request stops there and nothing is written. The automation calls the same service layer the interface calls, so there is one code path, not a side channel.
- Test and roll backI ship changes as pull requests through a merge queue that will not merge a failing branch. A failed smoke test or an SLO breach triggers an automated rollback, without waiting for me. The previous image stays deployable, so a rollback is a redeploy of a version that already ran.
Tools I use
| Tool | What it does here | Status |
|---|---|---|
| Luca public API (api.onluca.ai/v1) | OpenAPI 3.1 service with bearer keys, per-service scopes, 120 requests per minute per key and one response envelope. | Available |
| Agent tool bridge | Registers platform operations as model tools with declared parameters and required fields, and refuses unregistered ones. | Available |
| Multi-provider model gateway | One endpoint in front of several model providers, with scoped tokens, an allowed-apps list, rate limits and per-call logging. | Available |
| Luca MCP gateway | Routes MCP clients through the same audited tool endpoint the other surfaces already call. | Early access |
| Schema-validated call demo | Maps a plain-English request onto the published schema and rejects anything the schema does not describe. | Demo only |
Available means shipped and in use. Early access means built but not generally available. Roadmap means designed, not built. Demo only means it powers this public demo, not customer deployments.
Type a request in plain English and the demo maps it onto the published Luca API schema. It prints the endpoint, the scope and the body it would send, or the reason the schema rejects it. No call leaves the page.
Open the live demoThe other answers
LLM applications in productionEvery model call goes through one gateway, guardrails run on both sides, and each task gets a measured model choice.Prompt engineeringI write the output shape first, version the prompt, budget its tokens, and compare model replies before shipping.Retrieval-augmented generationGrounded, cited answers over the customer's own documents, with access rules applied before retrieval.AI agents and orchestrationI build agents that call your systems as registered tools, and the workflows that trigger and sequence them.Request classification and routingI build automation that classifies each inbound request, routes it by role, and holds risky changes for human approval.AI workflow orchestrationI design the steps, contracts, and data flows, then ship them as running workflows on customer systems.Hybrid enterprise integrationI connect AI services to your identity, applications and data, then run them where your policy allows.Prototype to productionI take a prototype through generated test suites and merge review gates to a zero-downtime deploy that rolls back on failure.Approvals and exception handlingI put a person between the agent and the change, with gated tools, a pending queue, expiry, and a record.Evals and output validationI score answers against a versioned rubric, trace failing calls to the integration that broke, and tune latency per call.Client-facing deliveryI turn engineering work into updates that client architects, security leads and business owners can each act on.
Gus IT LLC. Contact gus@gusit.de. Claude is a product of Anthropic PBC. Cloudflare and Workers AI are trademarks of Cloudflare, Inc. Llama is a trademark of Meta Platforms, Inc. Other names are trademarks of their respective owners. Gus IT LLC is an independent engineering firm and is not affiliated with or endorsed by these companies.