ProverCLI

A Python API and CLI for parsing outputs from the Certora Prover. This utility provides an easy-to-use interface for fetching and parsing prover verification results using either job URLs or job IDs.

Distributed on PyPI as certora-prover-cli; the Python import package is prover_output_utility (e.g. from prover_output_utility import ProverOutputAPI).

Documentation

📖 Full Sphinx documentation lives in docs/ — build it locally with make -C docs html (see docs/README.md).

The documentation includes:

• Complete API reference with all methods and classes
• Quick start guide and tutorials
• Usage examples and best practices
• CLI reference
• Authentication setup
• Caching documentation

Features

• Flexible Input: Accepts job URLs (format: /output/user_id/job_id), job IDs, and local emv-* folders
• Authentication: Interactive cookie-based login via certora_login; AWS SigV4 in CI; no auth for local folders
• Comprehensive Parsing: Extracts job info, verification results, rule details, calltraces, breadcrumbs, and statistics
• Error Handling: Robust error handling with custom exceptions
• Type Safety: Full type hints and typed dataclass results

Installation

As a CLI tool (recommended for end users)

Install via uv — each tool gets its own isolated venv, with one-command upgrade and uninstall:

# One-time uv install (skip if already installed)
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install from PyPI
uv tool install certora-prover-cli

# Upgrade / uninstall
uv tool upgrade certora-prover-cli
uv tool uninstall certora-prover-cli

Or with pip: pip install certora-prover-cli.

Local development install

git clone https://github.com/Certora/ProverCLI.git
cd ProverCLI
pip install -e .

As a dependency in another Python project

# pyproject.toml
dependencies = [
    "certora-prover-cli",
]

# requirements.txt
certora-prover-cli

Quick Start

Command Line Interface

After installation, you can use the prover-cli command:

# Get violated rules from a job (using job_id, not user_id)
prover-cli --job-id 67890
prover-cli --job-url "https://prover.certora.com/output/12345/67890/..."

# Save output to a file
prover-cli --job-id 67890 --output violations.json

# Get only job status
prover-cli --job-id 67890 --status-only

# Output as JSON
prover-cli --job-id 67890 --format json

Using as a Python Package

Data-fetching methods return typed dataclasses (see models.py), so results are accessed by attribute, not dict keys:

from prover_output_utility import ProverOutputAPI

# Initialize the API (interactive cookie-based auth via certora_login)
api = ProverOutputAPI()

# Accepts a job URL (format: /output/user_id/job_id), a job ID, or a local emv-* path
violations = api.get_violated_rules("67890")
for v in violations:                       # each v is a CheckResult
    print(f"{v.rule_name} ({v.method_name}): {v.assert_message}")
    if v.source_location:
        print(f"  at {v.source_location}")  # file:line

# Job status without fetching full output
status = api.get_job_status("67890")        # JobStatus enum
print(f"Job status: {status}")

Integration Example

from prover_output_utility import ProverOutputAPI
from prover_output_utility.exceptions import ProverAPIError


class VerificationAnalyzer:
    def __init__(self):
        self.api = ProverOutputAPI()

    def analyze_job(self, job_input):
        try:
            all_checks = self.api.get_all_checks(job_input)
            violations = [c for c in all_checks if c.is_violated]
            if violations:
                return {
                    "status": "failed",
                    "summary": f"{len(violations)}/{len(all_checks)} checks failed",
                    "violations": violations,
                }
            return {"status": "passed", "summary": f"All {len(all_checks)} checks passed"}
        except ProverAPIError as e:
            return {"status": "error", "error": str(e)}


result = VerificationAnalyzer().analyze_job("67890")  # job_id, not user_id
print(result)

Advanced Usage

# List recent jobs (returns JobInfo objects)
for job in api.list_recent_jobs(limit=5):
    print(f"Job {job.job_id}: {job.status} ({job.start_time})")

# Deep trace analysis for a violation
for v in api.get_violated_rules("67890"):
    if v.has_calltrace:
        trace = api.get_calltrace_for_violation("67890", v)
        print(f"{v.rule_name}: {trace.frame_count} calltrace frames")
    if v.has_breadcrumbs:
        bc = api.get_breadcrumbs_for_violation("67890", v)
        print(f"{v.rule_name}: {bc.total_steps} breadcrumb steps")

Authentication

Authentication is selected automatically by environment — see docs/authentication.rst for details:

api = ProverOutputAPI()                 # interactive / CI (auto-detected)
api = ProverOutputAPI(use_local=True)   # parse local emv-* folders, no auth

The legacy CERTORAKEY / certora_key= path is deprecated and no longer the recommended way to authenticate.

Output Structure

Prefer the typed accessors — get_violated_rules(), get_all_checks(), get_job_info(), get_calltrace(), get_breadcrumbs() — which return the dataclasses defined in models.py (CheckResult, JobInfo, CalltraceInfo, BreadcrumbInfo, …). The full API reference is in docs/. The lower-level fetch_output() returns the raw parsed dictionary for backward compatibility.

Error Handling

The API provides several custom exceptions:

• ProverAPIError: Base exception for API errors
• AuthenticationError: Authentication failed
• JobNotFoundError: Job not found or inaccessible
• ParseError: Failed to parse output
• InvalidJobError: Invalid job input format
• APITimeoutError: API request timed out

from prover_output_utility import ProverOutputAPI
from prover_output_utility.exceptions import JobNotFoundError, AuthenticationError

api = ProverOutputAPI()

try:
    result = api.fetch_output("invalid-job-id")
except JobNotFoundError:
    print("Job not found")
except AuthenticationError:
    print("Authentication failed - check your certora_login session")

Development

This repository holds the library/CLI code.

Code Quality

# Install development dependencies
pip install -e .[dev]

# Format code
black src/
isort src/

# Type-check
mypy src/prover_output_utility/ --ignore-missing-imports

License headers (REUSE)

uvx --from "reuse[charset-normalizer]" reuse lint

Building Documentation

ProverCLI uses Sphinx for generating official documentation:

# Install documentation dependencies
pip install sphinx sphinx-rtd-theme

# Build HTML documentation
make -C docs html

# View documentation
open docs/_build/html/index.html

The documentation includes:

• Complete API reference (auto-generated from docstrings)
• User guide and tutorials
• Usage examples
• Authentication setup
• CLI reference
• Caching documentation

See docs/README.md for detailed documentation build instructions.