Python — 144 Operations for AI Agents

This page is the canonical reference an AI coding agent uses to refactor, query, and analyze Python code through the act MCP server. 144 operations available: 70 refactor, 18 query, 42 analysis, 14 verification. Each operation is callable from Claude Code, Cursor, Codex, OpenCode, or any MCP-compatible agent host. Click any operation for a stable anchor link suitable for citation.

Worked Python examples

act101 follows Python's own conventions instead of imposing generic ones: a constructor derives from annotated class attributes, imports regroup into the standard-library / third-party / local order with the conventional blank lines, and guard wrapping produces a real try/except block at the surrounding indentation. Each example below is the verbatim result of the command shown, run against the file shown.

Generate __init__ from annotated attributes

`Reading` declares its attributes once, with annotations; the constructor should simply mirror them.

$ act refactor generate-constructor Reading

Before

class Reading:
    sensor_id: str
    celsius: float
    taken_at: str

After

class Reading:
    sensor_id: str
    celsius: float
    taken_at: str

    def __init__(self, sensor_id: str, celsius: float, taken_at: str) -> None:
        self.sensor_id = sensor_id
        self.celsius = celsius
        self.taken_at = taken_at

The annotations carry through to typed parameters and a `-> None` return — nothing is invented that the class did not already declare.

Regroup imports into stdlib, third-party, and local blocks

Five imports accumulated in arrival order, with a local model class stranded between standard-library modules.

$ act refactor import-organize imports.py

Before

from readings import Reading
import sys
import requests
import json
from collections import defaultdict

def load(path):
    return [Reading(**r) for r in json.load(open(path))]

After

from collections import defaultdict
import json
import sys

from readings import Reading
import requests

def load(path):
    return [Reading(**r) for r in json.load(open(path))]

`requests` lands alone in the third-party block and the local `readings` import drops to the bottom, each group separated by the conventional blank line.

Wrap the pipeline's crash site in try/except

`json.loads` on payloads from the field is where this parser actually fails in production.

$ act refactor wrap-try-catch --file parse.py --start-line 5 --start-column 5 --end-line 5 --end-column 27

Before

import json

def parse_reading(raw):
    data = json.loads(raw)
    if data.get("unit") == "F":
        celsius = (data["value"] - 32) * 5 / 9
    else:
        celsius = data["value"]
    return {"sensor": data["sensor_id"], "celsius": celsius}

After

import json

def parse_reading(raw):
    try:
        data = json.loads(raw)
    except Exception as e:
        raise

    if data.get("unit") == "F":
        celsius = (data["value"] - 32) * 5 / 9
    else:
        celsius = data["value"]
    return {"sensor": data["sensor_id"], "celsius": celsius}

The wrapped statement keeps the function's 4-space block indentation, and the default handler re-raises — the recovery policy stays with the author.

Query

18 query tools, the same on every supported language. Descriptions live in the shared reference: /docs/query-tools.

callers control_flow data_flow definition diagnostics effect_closure effect_summary fix_auto get_type graph import_organize interface mutations references repo_outline skeleton symbols symbols_batch

Refactor

Analysis

42 analysis tools, the same on every supported language. Descriptions live in the shared reference: /docs/analysis-tools.

analyze_api_diff analyze_chokepoints analyze_clones analyze_clusters analyze_cohesion analyze_conformance analyze_coupling analyze_cycle_risk analyze_cycles analyze_dead_code analyze_depth analyze_entry_points analyze_export analyze_extraction analyze_fan_balance analyze_features analyze_hotspots analyze_impact analyze_inconsistencies analyze_inheritance analyze_interface_bloat analyze_interfaces analyze_layers analyze_orphan_types analyze_patterns analyze_platform_deps analyze_readiness analyze_roles analyze_seams analyze_stability analyze_surface analyze_test_gaps analyze_thickness analyze_type_completeness churn_hotspots co_change_clusters coverage_overlay ownership_map profile_overlay simulate split_module trace_overlay

Verify

14 verify tools, the same on every supported language. Descriptions live in the shared reference: /docs/verification.

bisect_regression gate generate_test_harness scan secret_surface summarize_pr taint_flow unsafe_surface verify_behavioral_equivalence verify_contract_preserved verify_diff_semantics verify_port_parity verify_side_effects verify_test_impact