爱派 AiPy

Quick Verdict

AiPy is a local execution-oriented AI agent maintained by Beijing KnownSec. It connects model planning with Python, Node.js, and other executable code in a task, plan, code, execution, and feedback loop. Instead of merely returning a script, it can attempt to work with local files, browsers, desktop software, databases, and network resources to deliver a result. The official site remains active, the knownsec/aipyapp repository is not archived, its latest formal release is v0.5.0, and the vendor distributes Windows, macOS, and domestic Linux builds plus a commercial enterprise edition.

The benefit and danger come from the same property: model-generated code runs for real. Local execution does not guarantee offline processing. A configured model API, network script, browser automation, third-party package, points service, or enterprise integration may send prompts, code, file excerpts, or results to another provider. The public source is also not standard GPL-3.0: its license adds restrictions on unauthorized SaaS and integration into commercial products sold or given away. “Custom GPL-derived source-available license” is more accurate. Personal use has a free path, while points, model/API consumption, and enterprise delivery can cost money. Because of execution, licensing, and data-boundary risks, this directory does not recommend AiPy by default.

Best For

• Individuals and teams automating repeatable spreadsheet, public-data, document, and office tasks on a controlled computer.
• Technical users able to inspect execution traces, constrain directories and networking, review generated code, and recover from mistakes.
• Enterprises evaluating an intranet model, knowledge base, and local automation stack through a separate procurement and security review.
• Not suitable for unattended access to production servers, finance terminals, a primary personal machine, or environments containing broad credentials.
• Not suitable for developers requiring an OSI-standard open-source license or unrestricted commercial embedding and SaaS rights.

Key Features

• Task-to-execution loop: the model decomposes a request, generates code, runs it, observes output, corrects failures, and returns a result.
• Local resource access: can work with files, spreadsheets, databases, browsers, and desktop applications within operating-system permissions.
• Python ecosystem: generated code can use packages for analysis, visualization, network requests, document conversion, and automation.
• Desktop distribution: official downloads cover Windows, macOS, and environments such as Kylin and UnionTech UOS on supported domestic hardware.
• Model configuration: the public repository supports providers such as DeepSeek, while packaged products may use points or integrated provider channels.
• Enterprise edition: commercial private deployment adds knowledge bases, permission isolation, agent/skill production, auditing, and domestic model and compute integration.
• Task and Python modes: non-programmers can describe a desired result, while technical users can work more directly with Python execution.

Use Cases

Begin with a recoverable batch job. Copy non-sensitive spreadsheets into a dedicated directory, ask AiPy to produce a summary and charts, review the plan and code, and run against duplicates. Public-data collection, report drafting, and test-environment application control are other reasonable pilots. If the job is multi-model conversation without machine execution, Cherry Studio has a narrower permission surface. For repository-focused development, Claude Code or Cursor aligns more directly with version control and code review.

Do not let the model install arbitrary packages as an administrator, read an entire home directory, access browser cookies, SSH keys, cloud credentials, or production databases. Web pages, PDFs, emails, repositories, and API responses can contain prompt injection. Treat file writes, email sends, transactions, deletion, package installation, device control, and publication as high-risk actions that require the exact target and command to be displayed for human confirmation.

Pricing

The public repository’s latest formal release is v0.5.0, and individuals can download a client or install from source. The personal edition has a free path, but total cost can include several independent meters: AiPy points, model input and output tokens, embedding or multimodal APIs, paid data providers, third-party applications, and local compute and storage. Platform points do not necessarily cover model keys or external APIs configured by the user. Check each price and quota before running a large job.

The enterprise edition is quote-based and covers private deployment, intranet data, knowledge bases, permissions, auditing, training, custom agents, and model adaptation. Treat “data stays in the domain” as a deployment claim to verify against model location, egress policy, logs, telemetry, backups, updates, and remote support. It does not automatically apply to the personal edition or a configuration using cloud models.

Licensing needs special review. The repository license invokes GPL-3.0 and then adds two restrictions: unauthorized third-party SaaS is prohibited, and unauthorized integration into a commercial product for sale or giveaway is prohibited. These extra restrictions differ from standard GPL freedoms. Commercial hosting, OEM use, redistribution, or product integration requires written commercial permission from KnownSec.

Pros

• Real code execution can complete local work that a chat-only assistant cannot deliver.
• Chinese language, Windows, and domestic-system support reduce local adoption friction.
• A public repository and formal v0.5.0 release support implementation review and version pinning.
• Python provides broad coverage across data, documents, networks, and automation.
• The enterprise edition offers a procurement route for intranet deployment, permissions, auditing, and model integration.

Cons

• Generated code can read, modify, upload, or delete real data; errors and prompt injection have a large blast radius.
• “Local” is easily mistaken for offline, although cloud models, APIs, package downloads, and network scripts can exfiltrate data.
• The custom GPL-derived license restricts SaaS and commercial integration and is not standard GPL-3.0.
• Free access is not zero cost; points, model tokens, data APIs, and compute require separate budgets.
• Desktop automation and long tasks are sensitive to UI changes, dependencies, permissions, and model reliability.
• Enterprise security statements still require architectural verification, penetration testing, supply-chain review, and contractual commitments.

Alternatives

FAQ

Is AiPy still active?

Yes as of July 21, 2026. The official site distributes multiple desktop builds, the GitHub repository is not archived, the latest formal release is v0.5.0, and the enterprise edition launched in 2026.

Is AiPy completely free?

No. A personal free path exists, but points, model tokens, embeddings, multimodal services, external APIs, and compute can cost money. Enterprise deployment and services are quote-based.

Is AiPy standard GPL-3.0 open source?

No. Its license starts from GPL-3.0 but adds restrictions on unauthorized SaaS and commercial-product integration. It is more accurately described as custom GPL-derived source-available software.

Can local execution send data to the cloud?

Yes. Cloud models receive the context submitted to them, and generated scripts can call websites, APIs, or databases. Only a controlled architecture with local models, reviewed dependencies, restricted networking, and governed logs approaches offline operation.

Is generated code safe to run automatically?

Do not assume so. Run it in a restricted account, container, or VM; constrain directories and egress; inspect commands and package sources; and require confirmation for writes, deletion, installation, and external transmission.

How should an enterprise verify “data stays in the domain”?

Review model and embedding locations, DNS and egress, telemetry, logs, backups, remote support, update sources, and crash dumps. Confirm the result through contracts, testing, and runtime monitoring.

Bottom Line

AiPy differentiates itself by actually generating and running code so a natural-language task can end as a file, report, or computer action. Release v0.5.0, domestic desktop builds, and an enterprise edition show active development. Its execution authority, cloud-provider data paths, points and API costs, and non-standard GPL license cannot be reduced to “a local open-source agent.” Pilot in an isolated environment with duplicate data, pin the version, minimize permissions and egress, confirm high-risk actions, and obtain explicit commercial rights before restricted uses.