Custom browser agent + CapSolver Agent examples

Framework-agnostic examples for giving a custom browser agent the official capsolver-agent tool schemas and executor.

Examples only: no browseragent-capsolver package, framework fork, or duplicated SDK.

Repository scope

Use this repository when an agent runtime accepts JSON Schema or OpenAI-style function definitions but has no dedicated CapSolver adapter. The demo exports canonical schemas from the shared library and implements a minimal tool-call loop.

Quick start

git clone https://github.com/capsolver-ai/browseragent-capsolver.git
cd browseragent-capsolver
python -m venv .venv
source .venv/bin/activate
python -m pip install -r requirements.txt
playwright install chromium

Export .env.example values and run python examples/quickstart.py.

Key integration code

from capsolver_agent import create_executor, get_all_tools

tools = [tool.to_openai_function() for tool in get_all_tools()]
executor = create_executor()

# After the model requests a tool call:
result = await executor.execute(tool_name, tool_arguments)

See examples/quickstart.py for the complete request, execution, and response loop.

Project layout

examples/quickstart.py   Minimal schema-driven custom agent loop
requirements.txt         Shared SDK repositories plus OpenAI client
tests/test_demo.py        Offline validation
.github/workflows/ci.yml  Demo checks

Documentation

• CapSolver Agent tools
• CapSolver for AI agents
• CapSolver quick start

Responsible use

Use the example only for lawful, user-authorized workflows that respect target-site terms. Never commit secrets or private target data.

Contributing, support, and license

See CONTRIBUTING.md, SUPPORT.md, and SECURITY.md. Licensed under the ISC License.

The OpenAI client is used only as a sample function-calling runtime. This repository is maintained by CapSolver and is not affiliated with or endorsed by OpenAI.