AI operator
Maximo AI Chat, Voice, Files, Tools & Scheduled Work
Maximo AI is the business assistant inside MyTabulon. It can chat, use model-specific reasoning controls, work with attachments, generate and edit images, generate single or multi-speaker speech audio, create documents, run workspace tools, cite sources, create shares, organize sessions, run slash commands, accept steering context while it is working, schedule background AI work, use live voice when the plan allows it, and open from the official WhatsApp or Telegram bot links.
WhatsApp bot+234 808 721 6174Telegram bot@MaximoAIbot
Slash commands/side, /compact, /fork, /review, /goal
Maximo GoalsPlus: 1 goal, 50 turns; Pro: 1 goal, 200 turns; Max: 2 goals, 600 turns each
Computer commands/computer, /maximo (also /maximocode, /code), /codex, /syntax, /hermes, /openclaw
Slash contextActive Agent Skills and enabled MCP servers
Builder plugins@sites, @design, @forms, @notes
Running repliesSend more context without stopping Maximo
Computer modelThe selected Maximo chat model controls commands and final answer
Plan truthEffective owner account subscription for the active workspace
PersonalizationPersonality, Custom Instructions & About You; account-global
File upload limitPer file: 50 MB Free, 250 MB Plus, 2 GB Pro, 10 GB Max
Large uploadsDirect cloud upload when storage is configured
Cloud WorkspacePersistent project tree for later computer and agent runs
Side ChatSeparate, draggable, collapsible panel
In-app accessSidebar, chat start, Security
Guide searchSearch Maximo AI or slash commands
ReferralsGuide-backed answers; personal links and withdrawals
- Choose available models based on plan (DeepSeek, Maximo Pandora, Claude, GPT, Gemini, xAI Grok 4.6, Maximo Pandora image models, Grok Voice, and Gemini Live) or query custom Bring Your Own Model (BYOM) provider options when enabled. Grok Voice on paid workspaces uses grok-voice-think-fast-2.0 with High reasoning by default, 24 kHz realtime audio, improved live transcription and tool calling, and $0.08-per-minute usage accounting. Maximo Pandora 3.8 Pro Preview is available on Pro and Max for frontier coding, deep research, multimodal analysis, and complex agentic work; Pro gets up to 500K context tokens and Max gets the full 1M model window.
- Maximo Pandora 3.8 Pro Preview uses the shared cost-based AI pools at $0.80/M input and $4/M output through 226K input tokens, then $2/M input and $14/M output above 226K. Official Maximo benchmark results are coming soon.
- Maximo Atlas 1.2 is the current non-preview Atlas model on Plus, Pro, and Max with a 1M-token context window, 128K max output, and Low, Medium, High, X-High, or Max reasoning. It costs $0.55/M cache-miss input, $0.05/M cached input, and $1.50/M output. Maximo Pandora 3.8 is a Preview model on Plus, Pro, and Max with the same 1M context and 128K max output; it costs $1.50/M cache-miss input, $0.085/M cached input, and $4.20/M output.
- Maximo Pandora 3.8 Nano is now generally available on Plus, Pro, and Max; Atlas 1.1 is retained only as a legacy workspace model while Atlas 1.2 is the replacement.
- Maximo Atlas 1.1 leaves the Maximo AI provider on August 21, 2026. Keep using maximo-atlas-1.2 for new requests; existing Coding Plan Atlas 1.1 references already resolve to Atlas 1.2.
- DeepSeek's scheduled rates are selected deterministically from each request's UTC timestamp; when a live UI cannot know a future request's exact hour, it displays the current period and both published schedule branches instead of assuming peak or off-peak.
- Claude Opus 5 is available on Pro and Max for complex agentic coding, long-running work, deep reasoning, and professional tasks. It supports Low, Medium, High, X-High, and Max reasoning, up to 128K output tokens, 500K context on Pro, and the full 1M model window on Max. Usage accounting uses $5/M input, $0.50/M cached input, and $25/M output.
- When a managed model's standard input and output list prices add up to more than $15/M, Chat and Maximo Tag show a compact one-time notice above the composer that it can use shared AI usage pools faster. Got it remembers that model; Don't show again hides future high-cost notices on that browser.
- Managed chat models all draw from the same cost-based AI usage pools. More expensive or longer replies use more of those pools; there is no separate Grok 4.6 token quota. Pro uses up to 250K of Grok 4.6's 500K context window, while Max uses the full 500K. Grok 4.5 remains available as a legacy chat model.
- Maximo Atlas 1.2 and other leading coding models are also available outside the app through MyTabulon Coding Plans. Free starts with limited Atlas 1.2 access and no card. Plus is $3/week or ₦3,000/week, Pro is $9/week or ₦9,000/week, and Max is $15/week or ₦15,000/week. The gateway supports OpenAI-compatible Chat Completions and Responses APIs with full tool calling and no injected system prompt. Atlas Preview is removed from the Coding Plan on August 15, 2026 and replaced by Atlas 1.2.
- Bring Your Own Model (BYOM) allows your workspace to connect OpenAI, Anthropic, or OpenRouter keys under Settings > Integrations, making custom models available company-wide to all roles.
- Maximo reads the active workspace's effective owner-account subscription. If the workspace is covered by another owned workspace's billing home, Maximo should describe that inherited coverage instead of treating the workspace as Free.
- Use the public AI Benchmarks page to compare RPB, tool accuracy, reasoning, safety, sentiment, latency, cost efficiency, confidence, domain winners, and sanitized evidence across models.
- Reasoning effort controls balance speed and depth; preferences are saved by model.
- Personalization shapes how Maximo writes for you: choose a chat personality (Default, Friendly, Efficient, Professional, Candid, Quirky, Cynical, or Nerdy), add Custom Instructions, and fill in About You under Settings > Personalization. It is account-global and steers tone, style, and format only — never safety, your role, accuracy, tool rules, or your language. Turn it off anytime with the master switch.
- Attach supported images, PDFs, text files, audio, video, and other documents when the plan allows attachments.
- Type /side, /compact, /fork, or /review in the composer to open a side chat, compact the current context, create a new thread from this one, or ask Maximo to review the conversation, changes, mistakes, and missed requirements.
- Type /goal followed by what 'done' looks like to start Maximo Goals. Maximo keeps working toward that one objective across many turns, and a separate, low-cost evaluator checks after every turn whether the condition is met before the goal stops. Use /goal clear to stop a goal early.
- Goals run in the background and stream into the chat in real time when you reopen it; chats running a goal are marked with a goal badge in your history. Each goal-turn is a full Maximo reply that can use many tools, so a goal can run for hours and even span days via sleep/resume. Goals are available on paid plans: Plus runs 1 goal at a time with up to 50 turns, Pro 1 goal with up to 200 turns, and Max 2 concurrent goals with up to 600 turns each.
- Before a goal starts, Maximo checks your selected model's usage and warns you if it is low. If you continue and usage runs out during the goal, the run sleeps and resumes automatically when usage resets instead of failing. For any action that needs confirmation (refunds, payroll, moving money, sending invoices/emails), the goal pauses and asks; reply to confirm and it resumes.
- Type /computer, /maximo (or its aliases /maximocode and /code), /codex, /syntax, /hermes, or /openclaw when the next task should use an execution runner. /maximo is strict Maximo Code mode: Maximo must use the step-by-step Code Harness for executable coding work and cannot fall back to Cloud Computer or another agent. Maximo still stays the planner and final answerer; runners return real output for Maximo to use.
- Maximo Code now verifies changes through one language-agnostic code_verify engine. Fast edit checks target changed files; standard verification adds project type checks and focused tests; release verification can run full format, lint, test, build, and smoke commands before a pull request. Results include file, line, column, severity, rule, message, command, exit code, duration, and required-gate state.
- Installed TypeScript, Python, Rust, and Go language servers can contribute changed-file diagnostics without Maximo downloading extra tools. When no server is installed, the adapter is shown as skipped and the repository's normal verification commands still run.
- Maximo loads project guidance hierarchically from the repository root to the active file or package. MAXIMO.md, AGENTS.md, CLAUDE.md, .maximo/instructions.md, and path-scoped .maximo/instructions.json rules show their source, precedence, match scope, overrides, and provenance, and reload when work moves to another monorepo package.
- Lifecycle hooks can run at session start, before and after Bash, after edit/write/patch, before commit or pull request, and before stop. Managed and per-user hooks are trusted; project verification gates are declarative, while project shell commands require an explicit deployment trust policy or matching hash. Required commit, PR, and stop gates tell Maximo to continue fixing instead of shipping failed checks.
- The slash menu also shows active Agent Skills and enabled MCP servers. Select a skill or MCP server to attach it as context for the next request without turning it into a standalone command.
- Builder plugins extend Maximo with project tools. Type @sites to build, host, and publish websites and web apps with serverless backends and plan-based visitor file uploads automatically integrated with File Manager, @design to design, preview, and share decks, pages, prototypes, posters, and brand systems, @forms to build forms with conditional logic and analytics, or @notes to capture, organize, link, and act on notes and jots. A plugin's tools and guidelines are added to the model context only while it is active, so idle plugins cost zero tokens; viewers cannot activate builder plugins.
- Selected Agent Skill and MCP chips are hints, not permission overrides. Maximo still loads the selected skill or MCP server through the normal tools and only uses it when relevant.
- Maximo Cloud Computer is a tool, not a chat model. The selected Maximo model decides when to call it, writes the exact commands or script for built-in /computer runs, reads the returned output, and then answers.
- Use /computer for generic terminal work, browser or fetch inspection, website checks, Python/JavaScript/Bash execution, file outputs, screenshots, repository work, and project artifact capture. The built-in runner executes the model's commands; it does not use a hidden second AI planner.
- When a chat is attached to a project with Cloud Workspace enabled, Maximo Computer and configured Codex, Syntax, Hermes, or OpenClaw runs can restore the latest project workspace snapshot before work begins and save a new snapshot after durable changes.
- The Cloud Workspace is the same logical project tree across later runs, even when a fresh cloud runtime mounts it at a new absolute path.
- Files can open Workspace Files for the cloned repo tree and generated outputs, with a hidden-files toggle for system workspace folders and snapshots, plus download and IDE-style edit controls for text/code files.
- Use Codex, Maximo Syntax, Hermes, or OpenClaw only after an owner/admin connects the matching agent profile in Settings > Integrations.
- Codex profiles use Codex ChatGPT device login, with OpenAI API key as the fallback option.
- Maximo Syntax profiles use Maximo AI OAuth: click Connect, approve in the browser, and MyTabulon saves the encrypted profile when the callback returns.
- Side Chat keeps side questions out of the main thread. It opens a compact liquid-glass panel with its own icon and composer, supports collapse, expand, and drag, and can still be moved while collapsed on mobile.
- While Maximo is running, the composer stays available so the user can send more context, corrections, or steering without replacing the original request.
- Review prompts appear as compact review cards by default, with an expand control when someone needs to inspect the full prompt.
- Forked threads keep user and AI messages in chronological order so the new conversation remains readable.
- Use the official Maximo AI WhatsApp bot at +234 808 721 6174 or the Telegram bot at @MaximoAIbot when you want a direct messaging entry point outside the dashboard.
- The same bot links appear in the app sidebar, mobile menu, Maximo chat start screen, and Security settings under Bot Account Access.
- Ask Maximo to generate new AI images or edit existing image files from chat or voice. Outputs are saved to Files and returned as compact preview cards in chat.
- Ask Maximo to generate AI Audio (text-to-speech) from chat or voice with Maximo Pandora Voice. Choose any of 30 voices, single or up to two speakers each with their own voice and style, control emotion/accent/pace with style prompts and inline tags, and span 70+ languages. Audio (WAV) saves to Files and plays inline in chat. Voice Flash starts on Plus; Voice Pro starts on Pro.
- Ask Maximo to generate AI Docs from prompts and approved workspace files, then preview the result in chat, highlight selectable text, and send targeted revision instructions.
- Launch the AI Live Browser panel by asking Maximo to open a website, check a URL, or perform web research. Maximo can use its isolated cloud Chromium or, after you install MyTabulon Secure Browser, continue in a real tab from your own Chrome profile. The bridge is site-wide for normal HTTP(S) pages and is not hardcoded to a provider.
- For sign-in, choose You in the live browser. Maximo detaches from the tab and Chrome focuses it so you can enter passwords and verification codes privately. Return to MyTabulon and choose AI to continue in the same authenticated tab; credentials and Chrome cookies are never sent to Maximo.
- Opening the live browser does not replace or clear the active response. Partial answers, reasoning, tool events, and streaming remain in place, and the live-browser panel and activity pill stay attached only to the chat that created them.
- On desktop, drag the divider to resize the browser panel. Mobile keeps its dedicated responsive layout without the desktop resize handle.
- When Maximo saves outputs or uses File Manager records, normal file count, storage, and per-file upload limits still apply: Free 50 MB, Plus 250 MB, Pro 2 GB, and Max 10 GB for File Manager uploads.
- Large eligible File Manager uploads use direct cloud upload when cloud storage is configured.
- Maximo uses the signed-in user's role before choosing tools. Business Memory, sensitive company context, accounting, payroll, analytics, automations, company-wide AI action history, integrations, MCP, Agent Skills, team management, imports, refunds, credits, destructive deletes, and business settings edits are blocked for lower roles. Members can inspect, undo, and redo their own permitted Maximo AI actions from chat action history.
- Members cannot view or search company Business Memory through chat, voice, memory pages, background memory capture, or direct APIs.
- Member chat context is user-scoped for personal work. Maximo must not describe another teammate's leads, files, AI Docs jobs, or private memories as the current user's activity.
- Chat sessions and conversation history are user-owned. Switching to another team member account starts from that user's sessions, not a teammate's prior conversation.
- Viewers receive read-only answers for allowed areas and cannot ask Maximo to create, edit, delete, send, approve, export, connect, import, sync, run, refund, credit, schedule, generate AI Images, or revise AI Docs.
- Grok Voice keeps its own shared voice pool on paid plans and now uses Grok Voice Think Fast 2.0 with High reasoning by default; Gemini Live has a separate Pro/Max pool for live voice, camera/video, and attachment-aware sessions.
- Gemini Live limits are shown in AI usage: Pro gets 300 shared minutes monthly and 60 weekly per user, while Max gets 1,200 shared minutes monthly and 180 weekly per user.
- Chat and voice can create scheduled AI tasks for live event watching, research checks, follow-ups, reminders, and work that should start later or repeat.
- Scheduled AI tasks save every run inside a normal Maximo chat session so the user can read the full action trail later.
- Maximo can set first wake time, stop time, wake-every interval in minutes, max runs, notification channels, and can pause, resume, stop, delete, or run the task now.
- For referral questions, Maximo should use the guide and direct workspace users to Referral Earnings. Referral links, payout accounts, and withdrawals are individual to the signed-in user.
- Session tools include pin, rename, delete, branch from a message, move into projects, rate responses, and submit feedback.
- Shared chat links can expose a selected turn or conversation snapshot without giving dashboard access.
- Tool events show when Maximo reads, creates, updates, deletes, calculates, searches, or works across MyTabulon records.
- MyTabulon platform owners (the owner business workspace only) can run the Email Marketing Studio from chat: ask Maximo to design branded campaign emails from modular blocks, estimate an audience, send a test, then schedule or send. Campaign sends are high-risk actions that always show an Action Preview confirmation card, and the studio itself lives in the owner console at /dashboard/admin/email. These tools never appear in regular business workspaces.