Integuru v0

Generate Python integrations from browser network requests to internal platform APIs.

Stars
★ 4.8k
Last updated
4mo ago
License
AGPL-3.0
Primary language
Python

At a glance

Works with
Platform-specificOpenAI API
You'll need
Python 3.12PoetryOpenAI API keyShell / CLINetwork accessLocal filesystem
Typical use
A developer automating utility-bill downloads can capture one browser download flow and have Integuru trace the account and user identifiers needed by the request.
Main limitation
This is the earliest publicly released version; the repository directs users to integuru.com for the newer version.

What does this agent do, and when should you use it?

Integuru v0 is the publicly released early version of an integration agent that derives runnable code from browser network traffic. A user starts a browser with create_har.py, signs in, and performs the target action, producing request and cookie files. The integuru CLI then accepts a natural-language prompt and analyzes dynamic dependencies between captured requests. It can build a request dependency graph and generate Python code that executes the graph to perform the requested action. This version uses OpenAI cloud models, while the repository states that the newer Integuru version is available at integuru.com.

The user runs poetry run python create_har.py, logs into a destination platform in the spawned browser, and performs the desired action. The workflow creates network_requests.har and cookies.json, then runs poetry run integuru --prompt "download utility bills" --model <gpt-4o|o3-mini|o1|o1-mini>. Integuru identifies the target request, finds dynamic values such as accountId and userId, locates requests that provide those values, and attaches them into a dependency graph. It traverses from requests with no outgoing dependencies toward the target request and converts nodes into runnable functions. The CLI also exposes --har-path, --cookie-path, --max_steps, --input_variables, and --generate-code.

  1. A developer automating utility-bill downloads can capture one browser download flow and have Integuru trace the account and user identifiers needed by the request.
  2. An integration engineer working with a platform that lacks an official API can capture its browser traffic and generate Python calls to its internal endpoints.
  3. A developer who needs to break a browser action into dependent HTTP requests can use the generated graph to identify where dynamic request values originate.
  4. A developer exploring a graph that needs a value such as YEAR can supply key-value pairs through --input_variables.
  5. A maintainer validating the repository's behavior can run its documented test command, poetry run pytest.

How do you install or deploy this agent?

Use an OpenAI account and set the OPENAI_API_KEY environment variable. Install dependencies with poetry install, then enter the environment with poetry shell. For Jupyter use, run poetry run ipython kernel install --user --name=integuru. The documented CI workflow uses Python 3.12.

How do you use this agent?

Run poetry run python create_har.py, sign into the destination platform, and perform the target action to create ./network_requests.har and ./cookies.json. Then run poetry run integuru --prompt "download utility bills" --model gpt-4o. Use --har-path and --cookie-path for alternate input locations, and add --generate-code to request the full integration code. The repository recommends function-calling-capable gpt-4o for graph generation and says code generation automatically switches to o1-preview when that model is available in the user's OpenAI account.

What are this agent's strengths and limitations?

Pros
  • Uses captured browser requests and traces upstream sources for dynamic values such as accountId and userId, instead of treating a single recorded request as sufficient.
  • Produces a request dependency graph and can convert graph nodes into runnable Python functions and integration code.
  • Provides documented CLI controls for custom HAR and cookie paths, maximum steps, graph input variables, and full-code generation.
  • The repository states that captured data is stored locally in network_requests.har and cookies.json.
Limitations
  • This is the earliest publicly released version; the repository directs users to integuru.com for the newer version.
  • A user must log into the target site and perform the action in a browser; for 2FA sites, they must complete 2FA and obtain the relevant session material.
  • It requires an OpenAI API key and OpenAI cloud models including GPT-4o and o1-preview, with model availability dependent on the account.
  • Input variables are documented as graph-generation-only; code-generation support for input variables is described as forthcoming.

How does this agent compare with similar options?

Key facts side by side with the most closely related agents.

Agent Source review Stars Updated Language Full support on
Integuru v0 This agent 24 · Major gaps ★ 4.8k 4mo ago Python OpenAI API
Hello-Agents 54 · Major gaps ★ 81k 1d ago Python OpenAI API
Lagent Lightweight Agent Framework 43 · Major gaps ★ 2.3k 10d ago Python OpenAI API
OpenEvolve 40 · Major gaps ★ 7.4k 2mo ago Python Claude Code · OpenAI API

How does FollowAgents rate this agent?

FollowAgents source review · FARS-2.1
Major gaps
24/ 100 5-point scale 1.2 / 5
Trust 0/29
Reliability 3/14
Adaptability 6/18
Convention 7/18
Effectiveness 6/13
Verifiability 2/8
Why each dimension lost points
Trust0 / 29 · 0.0/5

Evidence shows: The repository provides no mechanisms for least privilege, user confirmation, data flow transparency, sensitive data handling, dependency security, external effects, rollback, or source attribution. README mentions local storage of HAR and cookies but does not address least privilege or user confirmation. Therefore all trust criteria score 0.

Reliability3 / 14 · 1.1/5

Evidence shows: README and pyproject.toml are consistent, but tests cover only partial functionality and no failure message handling is provided. Dependencies are listed in pyproject.toml but availability is not verified. Thus self_consistency scores 1, dependency_availability scores 1, failure_messages scores 0.

Adaptability6 / 18 · 1.7/5

Evidence shows: README describes target users (developers) and scenarios (reverse engineering APIs), but capability boundaries are not explicit. Triggering is via CLI arguments, but precision is not verified. Environment requires Python 3.12 and OpenAI API, but other environment fit is not described. Therefore each scores 1.

Convention7 / 18 · 1.9/5

Evidence shows: README provides installation and usage instructions, but lacks versioning/changelog and detailed known limitations. License is AGPL-3.0, but maintenance responsibility is not mentioned. Thus information_architecture scores 1, install_notes scores 2, naming_stability scores 1, examples_and_faq scores 1, known_limitations scores 1, license scores 2, versioning_changelog scores 0, maintenance_responsibility scores 1.

Effectiveness6 / 13 · 2.3/5

Evidence shows: Output is runnable Python code, but usability is not verified. Marginal value is high as it fills a gap for platforms without official APIs, but cost-benefit is not assessed. Therefore output_usability scores 1, marginal_value scores 2, cost_benefit scores 1.

Verifiability2 / 8 · 1.3/5

Evidence shows: Claims in README correspond to code, but lack external corroboration. Facts and inferences are not clearly separated. Thus claim_traceability scores 1, cross_source_corroboration scores 0, fact_inference_separation scores 1.

Risks and how to mitigate them
  • Not found in source: least-privilege scopingGrant only what the task needs: a dedicated account or read-only token, scoped to specific directories and repos.
  • Not found in source: confirmation before actingTurn on (or add) a confirmation step before it acts, and try it in a sandbox or test environment before real data.
  • Not found in source: data-flow disclosureWatch which external services it contacts (proxy or firewall logs) and keep sensitive data out until you know where it goes.
  • Not found in source: sensitive-data handlingUse dedicated, low-privilege, revocable API keys — never production credentials — and keep secrets out of logs.
  • Not found in source: dependency securityPin versions and run a dependency audit (npm audit, pip-audit) before installing; prefer running it in a container.
  • Not found in source: disclosed external effectsEstablish which external systems it writes to, sends to or changes, and verify with test accounts or repos before production.
  • Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
  • Not found in source: verifiable attributionInstall from the official repo or registry and check the publisher and URL to avoid look-alike packages.
  • The tool collects and stores user cookies and network request data, posing a risk of sensitive information leakage; ensure local storage security.
  • The tool relies on OpenAI cloud services, potentially sending data to third parties; users should be aware of privacy implications.
  • The generated code may violate the terms of service of target platforms; assess legal risks before use.
Evidence confidence: Low Reviewed Aug 09, 2026 Reviewed revision b063940a23de
See the full review method →

FAQ

Is it intended for platforms without official APIs?
Yes. Its documented approach is to derive code for a platform's internal endpoints from browser network requests, provided you can perform and capture the required browser action.
What credentials and files does it need?
It needs OPENAI_API_KEY. The workflow uses locally stored network_requests.har and cookies.json; for a 2FA destination site, complete 2FA and obtain the cookies, auth tokens, or session tokens first.
Can it generate executable integration code?
Yes. The README says it generates code to make all requests in the graph, and the CLI offers --generate-code for full integration code.
Which models are documented?
The CLI usage lists gpt-4o, o3-mini, o1, and o1-mini. The repository recommends gpt-4o for graph generation and says code generation can automatically switch to o1-preview when available.
View on GitHub ↗ Install ↓

Compare agents like this one

The same FARS review applied across the shortlist this agent qualifies for.

Related agents