Developing with AI agents

This article covers how to build on top of Fiber AI, whether with SDKs or your AI coding assistant of choice.

Overview

There are multiple ways to integrate Fiber AI into your workflow:

Start with the plugin if available for your IDE. It bundles everything — MCP servers, skills, and agent configuration — so you don't have to set up each piece manually. If a plugin isn't available for your IDE, you can install MCP servers, skills, and SDKs individually.

Prerequisites

Get Your API Key

1. Go to fiber.ai/app/api

2. Generate an API key (starts with sk_live_)

3. Set it as an environment variable:

export FIBER_API_KEY=sk_live_...

Add this to your shell profile (~/.zshrc or ~/.bashrc) so it persists across sessions.

Credits

Most Fiber AI operations consume credits. Credit costs vary by operation type — check the API documentation for current pricing.

• Check balance: fiber.ai/app/subscription

• Top up credits: fiber.ai/app/subscription

Plugins (Recommended)

Plugins are the fastest way to get started. A single install gives you everything — MCP servers, skills, and agent configuration — with no manual setup required.

What gets installed automatically

When you install the Fiber AI plugin, you get:

• 2 MCP servers — V2 (~10 direct tools for common operations) and Core (4 meta-tools for all 100+ API endpoints)

• 7 skills — search, enrich, audience, sdk-ts, sdk-py, setup, help

• Agent rules — credit awareness, authentication patterns, workflow best practices

• Session hooks (Claude Code) — reminds you to set your API key if it's missing

No additional configuration needed. Just install and start using.

Claude Code

claude plugin marketplace add fiber-ai/fiber-ai-plugin --scope project
claude plugin install fiber --scope project

Start Claude Code and run /fiber:help to see all available commands:

Cursor

Install from the Cursor Marketplace:

1. Open Cursor

2. Go to Settings > Plugins

3. Search for Fiber AI in the marketplace and install

Or install MCP only via deeplink — copy and paste into your browser:

V2 (recommended for most users):

cursor://anysphere.cursor-deeplink/mcp/install?name=fiber-ai-v2&config=eyJ0eXBlIjoiaHR0cCIsInVybCI6Imh0dHBzOi8vbWNwLmZpYmVyLmFpL21jcC92MiJ9

Core (all 100+ endpoints):

cursor://anysphere.cursor-deeplink/mcp/install?name=fiber-ai-core&config=eyJ0eXBlIjoiaHR0cCIsInVybCI6Imh0dHBzOi8vbWNwLmZpYmVyLmFpL21jcCJ9

Team Sharing

Commit an MCP config file to your repo so your entire team gets Fiber AI automatically — no individual setup required.

Claude Code — add .mcp.json at project root:

{
  "mcpServers": {
    "fiber-ai-v2": {
      "type": "http",
      "url": "https://mcp.fiber.ai/mcp/v2"
    },
    "fiber-ai-core": {
      "type": "http",
      "url": "https://mcp.fiber.ai/mcp"
    }
  }
}

Cursor — add .cursor/mcp.json (same format). VS Code — add .vscode/mcp.json (same format).

Individual Setup (If No Plugin Available)

If a plugin isn't available for your IDE, you can set up each component individually.

MCP Servers

MCP (Model Context Protocol) lets AI assistants use Fiber AI tools directly — search companies, enrich contacts, build audiences — all from your editor.

Fiber AI provides two MCP servers:

Both use HTTP (Streamable HTTP) transport. Most users install both for maximum flexibility.

For detailed setup instructions per editor (Claude Desktop, VS Code, Windsurf, Zed, Codex, and more), see Using Fiber via MCP.

Quick setup examples:

Claude Code (CLI):

claude mcp add --transport http fiber-ai-v2 https://mcp.fiber.ai/mcp/v2
claude mcp add --transport http fiber-ai-core https://mcp.fiber.ai/mcp

Codex:

codex mcp add fiber-ai-v2 --url https://mcp.fiber.ai/mcp/v2
codex mcp add fiber-ai-core --url https://mcp.fiber.ai/mcp

OpenCode

Add to opencode.json (project root, ~/.opencode.json, or $XDG_CONFIG_HOME/opencode/.opencode.json):

{
  "mcp": {
    "fiber-ai-v2": {
      "type": "remote",
      "url": "https://mcp.fiber.ai/mcp/v2"
    },
    "fiber-ai-core": {
      "type": "remote",
      "url": "https://mcp.fiber.ai/mcp"
    }
  }
}

OpenCode supports remote HTTP MCP servers natively — no mcp-remote bridge needed.For more see here.

VSCode

Add to .vscode/mcp.json:

{
  "mcpServers": {
    "fiber-ai-v2": { "type": "http", "url": "https://mcp.fiber.ai/mcp/v2" },
    "fiber-ai-core": { "type": "http", "url": "https://mcp.fiber.ai/mcp" }
  }
}

Any agent with stdio-only MCP support — use the mcp-remote bridge:

npx mcp-remote https://mcp.fiber.ai/mcp/v2

Skills

Skills teach AI agents how to use Fiber APIs correctly — proper authentication, parameter structure, workflow ordering, and error handling. They work with any agent that supports skills.sh (40+ agents including Claude Code, Cursor, Windsurf, Cline, and more).

npx skills add fiber-ai/fiber-ai-plugin --skill search
npx skills add fiber-ai/fiber-ai-plugin --skill enrich
npx skills add fiber-ai/fiber-ai-plugin --skill audience
npx skills add fiber-ai/fiber-ai-plugin --skill sdk-ts
npx skills add fiber-ai/fiber-ai-plugin --skill sdk-py
npx skills add fiber-ai/fiber-ai-plugin --skill setup
npx skills add fiber-ai/fiber-ai-plugin --skill help

SDKs

Use SDKs when you're building applications, scripts, or automations that call Fiber APIs programmatically. SDKs are for your code, not for your AI assistant — see MCP vs SDKs for when to use which.

TypeScript — 

npm install @fiberai/sdk
# or
bun add @fiberai/sdk

import { createClient, companySearch } from "@fiberai/sdk";

const client = createClient({ baseUrl: "https://api.fiber.ai" });

const result = await companySearch({
  client,
  body: {
    apiKey: process.env.FIBER_API_KEY!,
    searchParams: {
      industriesV2: { anyOf: ["Software"] },
      employeeCountV2: { lowerBoundExclusive: 50, upperBoundInclusive: 500 },
    },
    pageSize: 25,
  },
});

Python — 

pip install fiberai

import os
from fiberai import Client
from fiberai.api.search.company_search import sync as company_search_sync
from fiberai.models.company_search_body import CompanySearchBody

client = Client(base_url="https://api.fiber.ai")

result = company_search_sync(
    client=client,
    body=CompanySearchBody(
        api_key=os.environ["FIBER_API_KEY"],
        search_params={},
        page_size=25,
    ),
)

Full SDK documentation: https://docs.fiber.ai/article/fiber-sdks

Documentation Access via Context7

Fiber SDK packages (@fiberai/sdk on npm, fiberai on PyPI) are indexed in Context7, which means any AI agent with the Context7 MCP server can pull up-to-date Fiber SDK documentation automatically.

This is useful when your agent needs to look up specific function signatures, parameter types, or response shapes while generating code.

Claude Code:

claude mcp add --transport http context7 https://mcp.context7.com/mcp

Cursor — add to .cursor/mcp.json:

{
  "mcpServers": {
    "context7": { "type": "http", "url": "https://mcp.context7.com/mcp" }
  }
}

Once connected, your agent can resolve Fiber SDK documentation on demand when writing code.

Quickstart: Terminal Setup

Assume you have a fresh installation and want to start building with Fiber AI. Here are the commands for each agent.

Claude Code (Recommended)

# Set your API key
export FIBER_API_KEY=sk_live_...

# Install the Fiber AI plugin (auto-installs MCP + skills + hooks)
claude plugin marketplace add fiber-ai/fiber-ai-plugin --scope project
claude plugin install fiber --scope project

# Optional: add Context7 for SDK docs
claude mcp add --transport http context7 https://mcp.context7.com/mcp

# Start Claude Code
claude

Then try: /fiber:search "AI companies in San Francisco"

Codex

# Set your API key
export FIBER_API_KEY=sk_live_...

# Add Fiber MCP servers
codex mcp add fiber-ai-v2 --url https://mcp.fiber.ai/mcp/v2
codex mcp add fiber-ai-core --url https://mcp.fiber.ai/mcp

# Start Codex
codex

Then try: "Search for fintech companies in New York using Fiber AI"

Gemini CLI

# Set your API key
export FIBER_API_KEY=sk_live_...

# Add Fiber MCP servers 
gemini mcp add --transport http fiber-ai-v2 https://mcp.fiber.ai/mcp/v2
antigravity mcp add --transport http fiber-ai-core https://mcp.fiber.ai/mcp

# Start gemini
gemini

OpenCode

OpenCode

# Set your API key
export FIBER_API_KEY=sk_live_... #your api key

# Create opencode.json with Fiber MCP servers
cat > opencode.json << 'EOF'
{
  "mcp": {
    "fiber-ai-v2": {
      "type": "remote",
      "url": "https://mcp.fiber.ai/mcp/v2"
    },
    "fiber-ai-core": {
      "type": "remote",
      "url": "https://mcp.fiber.ai/mcp"
    }
  }
}
EOF

# Start OpenCode
opencode

Supported and Recommended Models

Fiber AI works with any AI model that supports tool use (function calling). Recommended models for the best experience:

Recommended

Also Supported

Any model that supports function calling / tool use will work with Fiber AI MCP servers. Performance may vary — models with stronger tool-use capabilities produce better results.

LLMs.txt

Fiber AI publishes an llms.txt file that provides a machine-readable summary of the API for AI agents. This is useful when:

• Your agent needs a quick overview of all available endpoints

• You want to give an LLM context about Fiber's API without installing an MCP server

• You're building a custom integration and want your LLM to understand the API surface

The llms.txt is also available as an MCP resource (llms://fiber.ai/llms.txt) through the Core MCP server.

MCP vs SDKs: When to Use Which

You can use both. For example, use MCP to explore and prototype searches interactively, then switch to the SDK to build the production version.

For detailed SDK documentation, see: https://docs.fiber.ai/article/fiber-sdks

For detailed MCP documentation, see: https://docs.fiber.ai/article/using-mcp-in-llms

Example Prompts to Get Started

Once your agent has Fiber AI connected (via plugin, MCP, or skills), try these prompts:

Search

• "Find AI companies in San Francisco with 50-200 employees"

• "Search for Series B fintech startups in Europe"

• "Find CTOs at SaaS companies in New York"

Enrich

• "Get the work email for linkedin.com/in/johndoe"

• "Enrich this company: stripe.com"

• "Find contact details for the CEO of Notion"

Audience (Bulk)

• "Build a list of 100 AI startups in the US and enrich their founders' emails"

• "Create an audience of Series A healthcare companies and export to CSV"

SDK Code Generation

• "Write a TypeScript script that searches for companies and saves results to a database"

• "Build a Python pipeline that enriches a list of LinkedIn URLs from a CSV file"

Account

• "How many credits do I have left?"

• "Check my Fiber AI connection status"

Rate Limits

Fiber AI enforces rate limits to ensure fair usage. If you hit a rate limit (HTTP 429), wait briefly and retry.

For current rate limit details, check the API documentation.

Best practices:

• Add retry logic with exponential backoff in production applications

• Use the audience workflow for bulk operations instead of making many individual API calls

• Check the response headers for rate limit information

Related Documentation

• Fiber AI — Main website

• API Documentation — Interactive API reference with schemas

• SDK Documentation — TypeScript and Python SDK guides

• Using Fiber via MCP — Detailed MCP setup per editor

• API Key Management — Generate and manage API keys

• Credits and Billing — Check balance and top up

• TypeScript SDK on npm

• Python SDK on PyPI

• Plugin Repository — Source for plugins, skills, and MCP configs