Ventures
AI agent cost control
A budget-control proxy for AI agents. Set a monthly cap per agent; once it is reached,
calls return 429 budget_exceeded before they reach the model
provider — so there are no partial bills and no overages to dispute.
- One-line integration: change
base_url, add two headers, keep everything else
- Multi-provider — Anthropic, OpenAI, Groq, Perplexity, xAI
- Smart Router: pass
"model": "auto" and the task shape selects the model
TypeScriptBunSQLiteStreaming proxy
Arabic-first legal operations · open source
A legal-operations platform for corporate teams reviewing bilingual Arabic–English B2B
commercial contracts. A real workflow — Client → Case → Document → AI review → human
approval → controlled export — not a legal chatbot.
- The data layer and the AI layer both resolve to mock or live from credential presence — one set of code paths serves the demo and production alike
- Retrieval filters on temporal validity and supersession; on a miss the model abstains rather than answering from unrelated corpus
- Deploys with no environment variables at all
Next.js 16React 19TypeScriptAnthropicSupabase
Streaming AI console with a hard budget gate · open source
The reference implementation of the AgentGuard stance: the live backend has
no direct-provider fallback. Every request carries budget credentials or it
does not go out at all.
- A provider that is not wired rejects explicitly rather than silently routing to another vendor
GET /api/health returns 200 only when every credential is present — and never echoes a secret
- Standalone Docker image; full mock mode for UI work with no keys
Next.jsTypeScriptDockerSSE streaming
Coortek
Early development
Scheduling infrastructure for high-performance teams · not yet released
An appointment-organisation layer built around the way teams that ship actually plan their weeks.
How I build
The constraint I hold across every product is that the demo path must always
work. A new engineer clones the repository, runs one command, and the whole
product is in front of them — seeded data, deterministic responses, every screen
reachable. No credentials, no setup call, no “works on my machine”.
That constraint is not a convenience. It forces every external dependency behind an
interface, which is precisely what makes the live path safe to change.
The second is that failure has to be explicit. A missing credential
returns 503, not a degraded answer. A provider that is not wired rejects, rather than
quietly routing somewhere else. A retrieval miss abstains, rather than answering from
whatever happened to be closest. Silent fallbacks are how systems lie to you in
production — and by the time you notice, the invoice has already been generated.
Questions
Who is Emanuele Cella?
Emanuele Cella is the founder and CEO of AgentGuard, a company building cost-control and safety infrastructure for AI agents. He also founded Lexai, an Arabic-first legal operations platform, and Agent Workspace, a streaming AI console with a mandatory server-side budget gate.
What is AgentGuard?
AgentGuard is a budget-control proxy for AI agents. You set a monthly spending cap per agent, and calls return 429 budget_exceeded once the cap is reached — before the request reaches the model provider, so there are no partial bills and no overages. Integration is one line: change the base URL and add two headers. It supports Anthropic, OpenAI, Groq, Perplexity and xAI.
What is Lexai?
Lexai is an Arabic-first legal operations platform for corporate legal teams reviewing bilingual Arabic–English B2B commercial contracts. It implements a Client → Case → Document → AI review → human approval → controlled export workflow, and is deliberately not a general legal chatbot.
What technology does he work in?
TypeScript across the stack, with Bun as the runtime and package manager, Next.js and React on the front end, and Anthropic and OpenAI models behind provider abstractions. Infrastructure work is streaming proxies, SQLite and Docker.
How do I get in touch?
Email cellaemanuele05@gmail.com, or open an issue on any repository at github.com/lele-ceo.