Personal and business AI assistant based on OpenClaw or ZeroClaw - full implementation of the existing protocol, multimodality, calendar/mail/calls, concierge, LinkedIn SMF, task management, and daily planning
Overview
It is necessary to create a Personal and business AI assistant that is fully implemented based on the already existing agent protocol OpenClaw or ZeroClaw. Important: we do not develop our own protocol and do not invent new orchestration rules - we take OpenClaw or ZeroClaw as a basis and build the entire system according to their principles, role structure, tool-calling approach, memory, guardrails, and agent lifecycle.
The assistant should not just be a chat-bot, but an operational execution layer: managing the calendar and tasks, preparing for calls, summarizing meetings, tracking agreements, supporting concierge service (travel/bookings) and LinkedIn SMF (content, likes, comments), as well as assisting with light product/site operations (structure, copywriting, work plan, release checklists).
Key protocol requirement (mandatory)
The system must be implemented strictly based on OpenClaw \ ZeroClaw or another fork
The performer must:
choose one of the protocols (OpenClaw or ZeroClaw) for MVP and fix this in the architecture,
demonstrate compliance with the implementation of the protocol (agent structure, memory outlines, tool rules, planning/execution logic),
not replace the protocol with a “similar” custom approach.
If the protocol has official repositories, documentation, or reference implementation, it should be used as a base.
Goals
Become a daily operational layer for personal and business execution.
Reduce time spent on planning, scheduling, mail, preparation for calls, and follow-ups.
Provide context from various sources: calendar, mail, calls/transcripts, tasks, notes.
Automate repetitive processes, with human confirmation for sensitive actions.
Scale to new features: LinkedIn SMF, concierge, product/site operations.
Main capabilities (Scope)
A. Implementation of OpenClaw/ZeroClaw as core
Fully implement the agent stack based on OpenClaw or ZeroClaw:
roles / agents according to the protocol,
lifecycle: planning - execution - reporting - memory,
tool-calling and access rules,
guardrails and security policies,
memory model according to the protocol.
Embed a “human approval layer” in those places where it is provided or allowed by the protocol, and connect approvals for:
B. Calendar
2-way integration with Google Calendar.
Creating/editing/canceling events.
Offering free slots, detecting conflicts.
Automatic buffers: focus blocks, time-to-travel, prep buffers.
Generating a daily plan based on tasks, goals, deadlines, and context.
C. Mail
Gmail integration.
Inbox triage: categorization, priorities, draft responses.
Extracting tasks/agreements from emails and placing them in the task system.
Follow-ups and reminders.
D. Calls and operational control
Importing transcripts/audio of calls (Zoom/Google Meet or uploads).
Summaries: decisions, action items, responsible parties, deadlines.
Auto-generating agendas for upcoming meetings based on previous agreements.
Tracking the execution of agreements and monitoring “what was agreed - what was done”.
E. Task management
For MVP choose one task tool: ClickUp
Creating tasks from mail/calls/notes.
Statuses, deadlines, overdue logic, weekly review.
Linking tasks with the calendar and daily plan.
F. Concierge (travel, bookings, services)
Finding hotels/flights, comparisons, recommendations.
Any final actions - only after approval.
Saving preferences (city, level, budget, loyalty, etc.).
G. LinkedIn SMF (Social Media Function)
Posting: preparing content, content plan, tone, categories.
Engagement: suggestions for comments/responses, like strategy.
Important: only “suggest + approve”, without aggressive automation that may violate LinkedIn policies.
Metrics: posting frequency, engagement, content categories.
H. Light product/site operations
Assistance in creating the structure of the site/product: sitemap, copy, backlog, release checklists.
Integrations with Webflow/GitHub - as a second stage.
Non-functional requirements
Security: encryption of secrets, least privilege, action auditing.
Reliability: retries, error handling, clear failure messages.
Observability: logs, tracing, dashboard for viewing agent runs and tool calls.
Privacy: retention control, ability to delete memory items.
Cost control: tracking model usage, rate limits, caching.
Delivery
Architectural document linked to OpenClaw/ZeroClaw (what and how is implemented according to the protocol).
Working MVP:
Documentation: setup, deployment, security, description of the protocol in the system.
Tests: core unit + integration smoke tests.
Deploy: basic CI/CD and deployment in the cloud.
What is required from the performer in response
Which protocol you choose for MVP: OpenClaw or ZeroClaw, and why.
Link to the repository/documentation of the protocol you will use.
How you will prove that the implementation complies with the protocol (mapping: roles, lifecycle, tools, memory, guardrails).
How you will implement approvals and audit log.
How you protect the system from prompt injection when reading mail/web data.
Estimation of hours and cost by milestones.
Examples of similar systems (links/demo/code).
Freelancehunt.
