Coding Agent

The coding agent works on a folder. It receives a task, reads and edits the files, runs your test command, then repeats until the tests pass or the iteration limit is reached. All file access is confined to the workspace you provide.

Tools

The agent calls a small toolset on your workspace:

Tool	Purpose
read_file	Read a file from the workspace.
write_file	Create or replace a file.
edit_file	Replace an exact block of text inside a file.
list_files	List the workspace files.
search	Search the workspace for a pattern.
run_bash	Run a command, for example the test suite.

Parameters

• api_key: The provider key.
• provider: Any chat provider, such as openai, anthropic or gemini.
• model: The model id, for example claude-sonnet-5 or gpt-5.5.
• workspace: The folder the agent is allowed to work in. Required.
• max_iterations: Cap on model turns, default 20.
• allow_bash: Set to False to disable command execution, default True.
• bash_timeout: Seconds allowed per command, default 120.
• log (optional): Print each tool call when True.

Example

from intelli.function.coding_agent import CodingAgent

agent = CodingAgent(
    api_key=YOUR_ANTHROPIC_KEY,
    provider="anthropic",
    model="claude-sonnet-5",
    workspace="./my_repo",
)

result = agent.run(
    "Fix the failing tests in calc.py",
    test_command="python -m pytest -q",
)

print(result["success"], result["summary"])

run() returns a dictionary with success, summary, iterations and test_output.

Inside a Flow

Use agent_type="coder". The task text is the coding task, and the output is a short summary that the next task can consume.

from intelli.flow import Agent, Task, Flow, TextTaskInput

coder = Agent(
    agent_type="coder",
    provider="anthropic",
    mission="fix the failing tests",
    model_params={
        "key": YOUR_ANTHROPIC_KEY,
        "model": "claude-sonnet-5",
        "workspace": "./my_repo",
        "test_command": "python -m pytest -q",
        "max_iterations": 20,
    },
)

fix_task = Task(TextTaskInput("The add function returns the wrong result."), coder, log=True)

flow = Flow(tasks={"fix": fix_task}, map_paths={"fix": []}, log=True)
result = asyncio.run(flow.start())