Bander

The description states that Bander is a system designed to allow family members (especially parents) to interact with AI assistants in a way that maintains control over what happens. It operates through two components:

• OpenClaw — the conversational agent that understands requests and selects bounded tools.
• Bander — a separate process that holds credentials, presents approvals via Telegram cards, and enforces user consent before executing anything.

Key features include:

• No direct access to real accounts by OpenClaw.
• All actions must be approved by the user in a visible card format.
• If conditions change between approval and execution (e.g., calendar event moved), Bander refuses to act.
• Uncertainty is handled explicitly — no false success or blind retries.
• The system uses GPT 5.6 Sol for language understanding, and integrates with Google Calendar, Gmail, and Telegram.

Not evidenced: actual product usage, customer feedback, or performance metrics beyond the author’s own tests.

The description states that Bander was built to answer one question: “What would an assistant have to look like for me to genuinely hand it to my parents?” This reflects a shift from general-purpose AI tools toward family-focused, safety-first design.

The author emphasizes:

• A focus on safety, especially around credential handling and user consent.
• An emphasis on transparency — showing exactly what will happen before doing it.
• A goal of making the assistant usable by people who are not technically savvy.

This positioning evolved from a desire to improve upon OpenClaw’s existing model, which the author believes lacks sufficient attention to safety for non-technical users. The project is framed as an experiment in building something safer and more trustworthy than current AI assistants.

Not evidenced: prior versions of the product, market positioning strategy, or competitive differentiation beyond self-description.

The description states that the target user is someone who wants to give a safe AI assistant to their parent or family member — particularly those who are not comfortable with complex digital tools. The author notes that his mother exemplifies this group.

Key characteristics of the intended user:

• Not technically proficient.
• Values control and transparency.
• Needs reassurance about what AI tools will do before acting.
• Likely to be older, possibly less familiar with modern software interfaces.

Not evidenced: specific customer segments, personas, or market size estimates. No mention of how many such users might exist or how they would be reached.

The description does not contain any information about pricing, monetization, or business model. It is unclear whether Bander intends to be a paid service, free-to-use, or part of a larger ecosystem.

Not evidenced: revenue streams, pricing tiers, subscription models, or commercial plans.

The author reports:

• Built using TypeScript, React, Node.js, and various OpenAI tools including GPT 5.6 Sol.
• Uses OAuth for secure credential handling.
• Implements a broker architecture with separate MCP tools for different integrations.
• Employs Codex for development and testing, including test-first loops.
• Includes a hosted browser experience to demonstrate behavior without account access.

Not evidenced: scalability, infrastructure details, deployment environments, or technical performance data beyond the author’s own tests.

The description states that:

• The project was submitted to the OpenAI 2026 hackathon.
• It includes a public repository with source code, setup guides, and test evidence.
• There are 535 functional cases and 26 adversarial cases documented.
• A hosted browser experience allows judges to observe deterministic outcomes.

However, there is no mention of:

• Real-world usage or adoption.
• Customer feedback or user testing beyond the author’s own.
• Product metrics such as active users, retention, or engagement.
• Any form of product launch or marketing effort.

Not evidenced: traction indicators, customer base, or market validation.

The description does not provide any information about competitors or similar products. It focuses solely on Bander's unique approach to safety and control, without referencing other tools in the AI assistant space.

Not evidenced: competitive landscape, existing solutions, or differentiation from current offerings.

Several risks are implied by the description:

• Limited scope: The system only supports Google Calendar, Gmail, and Telegram — limiting its utility.
• High setup friction: Requires 45 minutes to set up once per user, which may hinder adoption.
• No commercial viability: No indication of monetization or business model.
• Self-reported nature: All claims are unverified; no third-party validation or user data exists.
• Single-person team: Only one developer is involved, raising questions about scalability and long-term maintenance.

Not evidenced: risk assessments, failure rates, or mitigation strategies beyond the author’s own notes.

1. How many people have actually used Bander in a real-world setting?
2. What are the technical limitations of the current integrations? Are there plans to expand support for other platforms?
3. Can you describe the process of onboarding new users and how it might scale?
4. What is your plan for monetization or commercializing this product?
5. How do you intend to validate that Bander truly improves safety without reducing usability?
6. Have you considered how to handle edge cases in APIs that don’t support honest recovery?