Documentation
Make agents use fresh context.
Monte gives agents a persona-backed operating layer for the current task. The primary runtime surface is context, not a saved profile dump.
Install
Choose one package manager
Install the published CLI globally. The executable is always `monte`.
npm i -g monte-engineRun one install command per machine or environment. After install, run the guided setup wizard.
Guided quickstart
For most users, start with the setup wizard. It handles auth, source selection, optional agent skill creation, persona build, calibration, and a first context test.
Agent quickstart
Use the non-interactive path when an AI coding agent or workflow agent is bootstrapping Monte from files.
Sources and exports
Monte gets useful when it has evidence of how you actually decide, write, prioritize, recover from friction, and work with agents. Better sources produce better persona signals; thin or generic sources produce generic context.
Best sources
Chat exports, AI conversations, Obsidian vaults, Notion workspaces, long-form notes, decision logs, project docs, writing samples, task plans, retrospectives, and feedback you gave or received.
What Monte looks for
Repeated preferences, tradeoff patterns, communication style, friction points, execution habits, uncertainty, corrections, and source-specific contradictions.
Prepare exports
Prefer plain text, Markdown, JSON, CSV, PDF, Notion Markdown/CSV exports, Obsidian vault folders, or other document folders. Keep filenames meaningful and group sources by app or domain when possible.
Avoid low-signal data
Do not upload passwords, API keys, payment details, raw secrets, or files that do not describe how you think or work.
Guided sources
Build a persona from real sources
Run the guided setup wizard with a folder or export file. Monte previews useful files, uploads selected sources, queues persona build, and helps test the first context.
Hosted raw uploads are temporary processing inputs; retained evidence is normalized text, signal rows, and evidence packets.
Personalize context
Runtime
Get task-aware personalization
The primary runtime call. It returns the task-specific patch an agent should apply before answering.
Agents should read instructionBlock, personalizedMoves, executionProtocol, taskAntiPatterns, and truthSafety, then apply them silently.
Readiness
Check what to do next
A lightweight readiness/router call for agents that need to know whether to auth, build a persona, calibrate, or call context next.
Bootstrap is not the final personalization patch. Use context for real work.
Context modes
Use --mode when the task type is obvious. Modes make the returned moves and execution protocol fit the work instead of relying on broad auto-classification.
general
Default personalized help when the task type is mixed or unclear.
planning
Plans, sequencing, projects, events, constraints, and follow-through.
writing
Drafts, tone, messaging, content shape, and audience fit.
decision
Tradeoffs, recommendations, option scoring, and next moves.
learning
Explanations, teaching style, practice paths, and comprehension.
research
Source gathering, evidence synthesis, vendor scans, and comparisons.
creative
Naming, concepts, campaigns, stories, and divergent options.
coding/design
Preview modes. Use only when explicitly requested; not auto-selected from broad keywords.
monte personalize context "Help me approach this task" --mode general --jsonFiles and ingest
Async setup
Queue setup from files
Agent-friendly async setup. It checks auth, can create the Monte skill, analyzes files, uploads useful files, and queues persona build.
Use this when an agent is bootstrapping Monte from a folder of files. It returns before the hosted build is finished.
Poll
Check setup progress
Poll hosted setup readiness after an async setup call.
Keep polling until persona readiness is clear, then call personalize context for the actual task.
Preview
Inspect files before ingest
Preview which local files Monte would include or skip before upload.
Dry-run is useful before uploading large exports or mixed folders.
Agent protocol
Agents should use Monte before planning, writing, tradeoffs, recommendations, research, creative work, and decision support. They should skip Monte for trivial facts, simple calculations, and formatting-only edits.
Agent command
Default agent call
Use this before work where priorities, tradeoffs, taste, writing style, or follow-through matter.
Do not call Monte for trivial facts, simple calculations, or formatting-only edits.
Read the JSON and apply instructionBlock, personalizedMoves, executionProtocol, taskAntiPatterns, and truthSafety silently. Pick an explicit --mode for clear planning, writing, decision, learning, research, or creative work.
Agent skill file
The skill file is the repeat-usage layer. It tells an agent when to call Monte, when to skip it, and why fresh context beats a saved profile dump.
Agent skill
Generate a reusable skill
Create or verify <dir>/Monte/SKILL.md so an agent knows when to call Monte, which modes to pass, and when to skip it.
Do not assume the target directory; every agent loads skills from a different place.
Manual auth
monte setup handles auth for most users. Use these commands when you only need to inspect or repair CLI authentication.
Authenticate
Store your Monte API key
Prompt for a hosted Monte API key, verify it against the runtime API, and store it in ~/.monte/config.json.
You do not need to set the hosted API URL manually.
Status
Verify auth state
Check which API URL and key source the CLI is using before an agent starts important work.
Useful when a shell, agent, or demo machine might be using a different local config.
Profile export is not runtime
monte personalize profile --json is an inspect/export snapshot, not the main agent runtime. It can be saved locally, but it goes stale, is not task-scoped, lacks live calibration and usage controls, and should not replace fresh monte personalize context calls for important work.
Use profile JSON for debugging, inspection, export, or verifying that a persona exists. Use context when an agent is about to do real work.
CLI reference
Persona inspection
Use these to inspect readiness and persona state. They are not a replacement for task-scoped context.
Persona
Check persona status
Show whether a persona exists and whether the account is using a real persona or a demo fallback.
Inspect
Inspect profile JSON
Export a broad profile snapshot for debugging, inspection, or export.
This snapshot can go stale, is not task-scoped, and should not replace fresh context calls.
Calibrate
Tune persona assumptions
Answer or revise high-impact inferred persona items so corrections survive future rebuilds.
Diagnostics
Use these when the CLI environment or runtime connection looks wrong.
Doctor
Diagnose CLI state
Run local CLI diagnostics for config, runtime reachability, and common setup issues.
Config
Show config
Inspect the CLI config stored under ~/.monte/config.json.
The hosted CLI already defaults to https://api.monteengine.com. Use set-api only for local/dev or alternate deployments.
The hosted CLI already defaults to https://api.monteengine.com. Use monte config set-api only for local development or alternate deployments.
Pricing
Hosted billing is prepaid wallet-based. Real-persona context calls cost $0.20 each. Demo-persona context calls are free up to the demo cap.
Markdown and LLM docs
Troubleshooting
Missing API key
Run monte setup for guided setup or monte auth for manual auth repair.
No persona
Run monte setup with useful sources or use monte setup --files <path> for agent setup.
Weak personalization
Add higher-signal sources: conversations, decisions, plans, notes, writing samples, feedback, and project docs.
No credits
Open dashboard billing and top up. Real-persona runtime calls return HTTP 402 without credits.
Saved profile feels wrong
Regenerate task context. Profile exports are stale snapshots and are not task-scoped.
npm install warnings
Transitive npm deprecation warnings are non-blocking if monte doctor --agent --json passes.