MCPWorks with Claude Code, Codex

One API for every tool your AI agent needs.

Search thousands of APIs, call them through one endpoint and pay only for what your agent actually uses.

  • No subscriptions
  • No API-key jungle
  • Pay per call

Start with a $20 balance — it moves only when your agent calls a tool.

agent session
demo

Task

Find companies that raised Series A in the last 90 days and identify their marketing leads.

Searching tools…

Illustrative demo. Tools, prices and results shown here are mock data.

The problem

Your agent shouldn’t need 30 subscriptions.

Each capability is another signup, another invoice, one more key to rotate. All that wiring — and the agent still can’t use a tool nobody set up in advance.

Today

9 keys, 9 invoices

AI Agent
  • SEMrush key
  • Hunter key
  • SERP API key
  • Scraping API key
  • Social API key
  • Company API key
  • Maps API key
  • Another API key
  • Another subscription

With one API

1 key, 1 balance

AI Agent
ONE API
Any tool it needs
  • SERP
  • Backlinks
  • Keywords
  • Scraping
  • Social
  • Company
  • People
  • Emails
  • Maps
  • Finance
  • OCR
  • Translation

Discovered at runtime, billed per call.

A reference point, if you want one: OpenRouter, but for tools instead of models.

How it works

Three steps, no integration work.

Discovery, pricing and execution run through one connection, so an agent can adopt a tool it has never seen before — mid-task, without a deploy.

  1. 01

    Search

    The agent describes what it needs in plain language. No catalog to memorize, no integration to write in advance.

    search("find backlinks for domain")
  2. 02

    Select

    It gets back matching tools with their input schema, latency and exact price per call — enough to choose on its own.

    → seo.backlinks · $0.01 / call
  3. 03

    Call

    One endpoint, one key. The balance is charged only for calls that return a result.

    call("seo.backlinks", { target })

Tool explorer

Ask for a capability, not for a vendor.

This is what your agent sees when it searches: matching tools, the provider behind each one, the schema it can call and the price it will pay.

6 of 27 capabilities in the demo catalog

  • Google search results

    Serplane · SERP

    $0.002 / call

    Live organic results, ads and related queries for any keyword and location.

    ~1.4 s
  • Work email lookup

    Mailtrace · Email discovery

    $0.008 / call

    Find the most likely work email for a person at a given company domain.

    ~1.0 s
  • Funding rounds

    Firmly · Company intelligence

    $0.012 / call

    Rounds, amounts, investors and dates, filterable by stage and window.

    ~890 ms
  • Web page scraping

    Fetchlayer · Web scraping

    $0.003 / call

    Render a page with a headless browser and return clean text, HTML or markdown.

    ~2.6 s
  • Backlink profile

    Linkgraph · Backlinks

    $0.01 / call

    Referring domains, anchors and authority scores for a domain or URL.

    ~2.1 s
  • People enrichment

    Roleframe · People enrichment

    $0.006 / call

    Decision makers at a company with title, seniority and department.

    ~710 ms

Use cases

One task, many tools, no wiring.

Chains an agent runs end to end — without you opening a single vendor account.

  • Market research

    1. Competitors
    2. traffic
    3. ads
    4. keywords
    5. backlinks
    6. decision makers
  • Sales agent

    1. Company
    2. people
    3. email
    4. verification
    5. CRM
  • SEO agent

    1. SERP
    2. keywords
    3. backlinks
    4. competitors
    5. content opportunities
  • Growth agent

    1. Ads
    2. creatives
    3. social
    4. traffic
    5. market signals
  • Coding agent

    1. Search for a capability
    2. call external API
    3. continue working

Pricing

Pay per call. That’s the entire model.

Agents don’t spend like people. They burst — heavy one day, silent the next. Subscriptions bill for the idle days; a balance doesn’t. Top it up, spend it call by call, cancel nothing.

Balance

$20to start

Credits sit on one balance and are drawn down call by call, across every provider in the catalog.

  • No monthly fee, no seats, no minimum spend.
  • Failed calls aren’t billed.
  • Every call returns its exact cost in the response.
Get API key

What $20 buys

  • Search results

    from $0.002 / call

    10,000 calls
  • Email verification

    from $0.003 / call

    6,600 calls
  • Keyword data

    from $0.006 / call

    3,300 calls
  • Company enrichment

    from $0.01 / call

    2,000 calls
  • Backlinks

    from $0.01 / call

    2,000 calls

Demo pricing. The live catalog will publish each provider’s real per-call price before an agent can spend anything.

For developers

Integration is one install and one key.

One package for code, one command for MCP. Nothing to configure per provider, nothing to redeploy when your agent needs a new capability.

You integrate once. Every capability after that is zero engineering: the agent finds it, reads the schema and calls it.

REST or SDK
Same endpoint for discovery and execution, typed responses, cost returned with every call.
MCP server
Install once and Claude Code, Codex or your own agent can search the catalog on its own.
Install
npm install apiviaai
Use
import { tools } from "apiviaai";

const [tool] = await tools.search(
  "find work email for a person"
);

const result = await tools.call(tool.id, {
  full_name: "Dana Whitfield",
  domain: "example.com",
});
MCP
apiviaai mcp install

Security

Your secrets stay secret.

An agent that can call anything shouldn’t also be holding every credential you own. Here is how the platform is being designed.

Planned architecture — not shipped yet

  • Bring your own provider key

    If you already pay a provider directly, the key will live on the gateway side — attached to your account, not to a prompt.

  • Agents get results, not credentials

    The agent asks for a capability and receives a response. The raw secret is never part of the tool output it can read or log.

  • Scoped keys and spend limits

    Separate keys per agent, per-call and per-day ceilings, and a full log of what was called and what it cost.

FAQ

The model, in five answers.

  • What am I paying for?

    Completed calls, nothing else. Every tool lists its price up front, and the exact amount comes back in the response. No subscription, no seats, no minimum.

  • What happens when the balance runs out?

    Calls stop. Nothing renews and nothing keeps charging — top up when you’re ready.

  • What if a call fails?

    You aren’t billed. An error returns an error, not a charge.

  • Can I cap what an agent spends?

    Yes — separate keys per agent, per-call and per-day ceilings, and a log of every call with its cost. Spend controls are part of the platform design, not an add-on.

  • Do I need accounts with the providers?

    No. Providers sit behind the gateway. Your agent sees the capability, the schema and the price — you see one balance and one invoice.

Give your agent the internet’s toolset.

One connection. Thousands of capabilities. Pay only for what gets used.