emprise

The AI partner that runs on your terms — any model, your machine, your data.

brew install senzalldev/tap/emprise                 # CLI
brew install --cask senzalldev/tap/emprise-desktop  # Desktop (macOS)
curl -fsSL https://emprise.dev/install.sh | sh

Windows & Desktop downloads · then run emprise init in any repo and paste an API key — that's the whole setup.

A CLI and matching Desktop app: OpenAI, Anthropic, Google, or fully local with Ollama. 96+ tools, vision paste, document intelligence, PHI/HIPAA mode, undo for every file edit, and plugins from any shell script. One config, one history, both surfaces.

⚠ Experimental — work in progress, no warranty. Source is private; we ship signed builds. Use at your own risk.

Why emprise Wiki
96+Tools
22Categories
6LLM Providers
338Tests

The idea

In 1960, J.C.R. Licklider described man–computer symbiosis: machines handling "the routinizable, clerical operations that fill the intervals between decisions," so people can think. emprise is that idea, shipped — an AI partner that runs the commands, reads the files, and watches the output while you set the goals and make the calls.

It runs on your terms: pick OpenAI, Anthropic, or Google — or stay entirely local with Ollama and let nothing leave your machine. Every file edit is checkpointed with undo. Costs are shown in real dollars. Your scripts become its tools. A partner, not a parrot.

Desktop App, at a glance

Native macOS / Linux / Windows app built with Wails v2 + Svelte. Same Go runtime as the CLI, same config, same SQLite history.

Emprise Desktop welcome screen
Welcome screen with attached-context picker (Index / DB / MCP / Web / Memory)
Profiles tab — 28 configured
Profiles — Detect Ollama, GitHub Models OAuth, Azure, custom OAuth all in one place
Prompts tab with emphasis categories
The 6-layer prompt system with toggleable emphasis categories
Data sources — indexes, web contexts, databases
Data sources — FTS5 document indexes, web contexts, SQL databases (incl. Azure AD)
Compose prompt modal with AI-refined version
Prompt composer — AI-refined version side-by-side with your draft
Tool stream produces a PDF artifact with Open folder action
Generated artifacts — click the path or Open folder to launch in Preview / default app

Features

Everything you need to work with AI assistants, manage your environment, and handle documents — from the terminal or the desktop.

96+ Built-in Tools

System monitoring, file management, git operations, shell execution, Docker, SSH, databases, web fetching, document parsing, image handling, JSON processing, and more — across 22 categories.

Document Intelligence

Index documents with FTS5 full-text search. Generate PDF, DOCX, HTML, Markdown, and plain text. Parse and ingest PDF, DOCX, XLSX, CSV, and more into conversations.

7 LLM Providers

Ollama (local), OpenAI, Anthropic, Google AI, GitHub Models, Azure OpenAI / AI Foundry, and custom OAuth endpoints. Switch providers instantly with profiles.

Vision: Paste Images

Paste an image from your clipboard — terminal or desktop — and the model actually sees it. Multi-image attachments with automatic downscaling, across OpenAI, Anthropic, and local multimodal models.

Reliable & Cost-Smart

Automatic retries with visible backoff status, provider prompt caching (99% cached input measured on repeat turns), lazy tool loading that cuts per-round tokens ~74%, and real-dollar session costs with zero setup.

Instant Key Setup

Paste an API key — emprise detects the provider, creates a tested profile in one keystroke, and never sends the key to a model. OpenAI, Anthropic, and Google keys recognized on sight.

Built for Scripting

emprise --json emits machine-readable results with exit codes for cron and CI. Search your chat history with /find, attach files with @path and Tab completion.

PHI/HIPAA Compliance

Toggle PHI mode to restrict all data to local-only processing. Enforces Ollama-only, disables network tools, enables audit logging, validates endpoints.

6-Layer Prompt System

Base identity, capability tier, role defaults, user rules, model overrides, and emphasis rules. Role presets for developer, devops, cloud, data, security, and admin.

Database Tools

Query SQLite, PostgreSQL, MySQL, and SQL Server. Inspect schemas, describe tables, convert CSVs to SQLite. Formatted table output with execution timing.

Desktop + CLI

Native desktop app built with Wails + Svelte alongside the terminal CLI. Both share the same configuration, tools, prompts, and conversation history.

OAuth 2.0 Enterprise

Device-code and authorization-code flows for enterprise identity providers. Automatic token refresh. Connect to any OpenAI-compatible endpoint behind OAuth.

Cron Scheduling

Schedule recurring tasks with cron expressions. Automated queries, system checks, report generation, and more — running on your schedule.

Config Versioning

Configuration backup and versioning with export/import. Share setups across machines. Track changes to your emprise configuration over time.

Configurable Output

Set output limits per model or globally. Control response length, tool calls per turn, and context window usage. Tune for speed or depth.

Dev Environment Audit

/dev scans 40+ tools across categories. Missing something? Select it and emprise installs it. Platform-aware: brew, apt, or winget.

Desktop + CLI

Two interfaces, one brain. The Desktop app (Wails + Svelte) and the CLI share the same config, tools, prompts, and conversation history.

Desktop App

  • Native window with Wails + Svelte
  • Dark, Light, and OLED themes
  • Visual tool management
  • Conversation browser with search
  • Document viewer and indexer
  • Profile switching with UI

CLI

  • Full TUI with arrow-key menus
  • Pipe mode: cat file | emprise
  • Single-shot: emprise "question"
  • Tab completion for all commands
  • Diff view with syntax highlighting
  • Background processes and streaming

Both apps read from ~/.emprise/config.yaml — change a setting in one and it applies to both.

emprise
/ | \ emprise v0.3.0 '--+--' anthropic (claude-sonnet-4) | ~/project ──────────────────────────────────────────────────── > Ask anything... (/help for commands) ──────────────────────────────────────────────────── /help · esc · ctrl+c · ctx:0% · ~/project

96+ Tools in 22 Categories

All tools execute asynchronously. Toggle on/off per category with /tools. Every tool call is visible with arguments and output.

System

get_os, get_cpu, get_memory, get_disk, get_processes, get_network, get_env, get_uptime

Files

read_file, write_file, list_dir, search_files, grep_files, edit_file, create_file, append_file, move_file, copy_file

Git

git_status, git_log, git_diff, git_branch, git_blame, git_pr_diff, git_commit, git_stash, git_tag

Shell

run_command — streaming output, configurable timeout, safety blocklist

Web

web_search, web_fetch, web_scrape

Docker

docker_ps, docker_logs, docker_images, docker_compose_status, docker_exec

Process

start_process, check_process, stop_process, list_processes_managed

SSH

ssh_run — connect to configured remote hosts

Clipboard

clipboard_read, clipboard_write — pbcopy/xclip/clip

Dev

run_tests, check_dependencies, notify, lint, format

Image

read_image, screenshot, image_resize

JSON

json_query, json_schema, json_transform — parse, query (jq-style), extract schemas, transform

SQL

sql_query, sql_schema, sql_describe, csv_to_sqlite — SQLite, PostgreSQL, MySQL, SQL Server

Document Parsing

read_folder, read_pdf, read_docx, read_xlsx, read_csv — batch directory ingestion

Document Generation

generate_pdf, generate_docx, generate_html, generate_markdown, generate_text

Document Index

index_documents, search_index, index_status — FTS5 full-text search engine

Scheduling

cron_create, cron_list, cron_delete, cron_status — recurring task automation

Config

config_export, config_import, config_backup, config_restore, config_diff

Network

http_request, ping, dns_lookup, port_scan

Archive

compress, decompress, list_archive

Math

calculate, convert_units, statistics

Batch

batch_run, parallel_exec — run multiple operations concurrently

Document Intelligence

Index, search, generate, and parse documents — all from the same interface.

FTS5 Search Index

Index documents into a local FTS5-powered search engine. Full-text search across your document library with ranked results, snippets, and metadata filtering.

Document Generation

Generate documents in five formats: PDF, DOCX, HTML, Markdown, and plain text. Ask the LLM to create reports, summaries, or documentation and export directly to file.

Document Parsing

Feed PDF, DOCX, XLSX, CSV, and plain text into conversations. Batch-ingest entire directories. Content is extracted and sent as LLM context automatically.

document indexing
> Index all the docs in ./contracts and search for renewal terms ● index_documents(./contracts) Indexed 47 documents (PDF, DOCX, TXT) FTS5 index: 128,340 tokens ● search_index("renewal terms") ┌────┬──────────────────────────┬───────┬──────────┐ │ # │ Document │ Score │ Format │ ├────┼──────────────────────────┼───────┼──────────┤ │ 1 │ vendor-agreement.pdf │ 0.94 │ PDF │ │ 2 │ service-contract.docx │ 0.87 │ DOCX │ │ 3 │ lease-addendum.pdf │ 0.72 │ PDF │ └────┴──────────────────────────┴───────┴──────────┘ Found 3 documents with renewal clauses. The vendor agreement has an auto-renewal on Jan 1 with 60-day notice required...

LLM Providers

Connect to any provider. Switch between them instantly with /profile.

Ollama

Local / Free

Run models locally with complete privacy. No API key, no network. Auto-detects available models. Required for PHI compliance mode.

OpenAI

Cloud / API Key

GPT-4o, GPT-4o-mini, o1, o3, and all OpenAI models. Pay-per-token pricing.

Anthropic

Cloud / API Key

Claude Opus, Sonnet, Haiku. Pay-per-token. Excellent for code review and analysis.

GitHub Models

Cloud / Free Tier

Access models through your GitHub subscription. Uses your existing GitHub token.

Azure OpenAI

Cloud / Enterprise

Connect to Azure-hosted OpenAI deployments. Enterprise compliance and data residency.

Custom OAuth

Enterprise / OAuth 2.0

Any OpenAI-compatible endpoint behind OAuth. Device-code and authorization-code flows with automatic token refresh.

6-Layer Prompt Engineering

emprise assembles the system prompt in 6 layers. Each layer adds context and rules that shape LLM behavior. Full docs on the wiki.

1

Base Identity

System info, OS, working directory, platform tools, emprise version

2

Capability Tier

Auto-detected from model name. Small models get simpler instructions, large models get full context.

3

Role Defaults

6 presets: developer, devops, cloud, data, security, admin — each with domain-specific rules

4

User Rules

Custom instructions, user name, project context from .emprise.md files

5

Model Overrides

Per-model tool limits, response length caps, extra rules for specific models

6

Emphasis Rules

40+ toggleable rules. High emphasis repeats 3x for models that tend to ignore instructions.

/prompts
Prompt Emphasis Role developer Name Steve LLM Tier advanced > Code Practices 5/7 active Developer Workflow 3/4 active DevOps / Cloud all off Security 3/6 active Test prompt ← Back
config.yaml
prompts: user_name: Steve role: developer emphasis: write_to_file: high branch_workflow: medium docker_security: off model_overrides: "llama3.2:3b": max_tools_per_turn: 1 extra_rules: - "respond concisely"

PHI/HIPAA Compliance

Toggle PHI mode for environments where protected health information must never leave the device.

Data Isolation

PHI mode enforces local-only LLM providers (Ollama). All clipboard, web, and SSH tools are disabled. Data never leaves your machine.

Audit Logging

Every operation is logged to a dedicated audit file. Track what was queried, when, and which tools were invoked. Full accountability trail.

Endpoint Validation

emprise validates that configured endpoints are truly local before allowing any data transmission. Prevents accidental cloud routing.

One-Command Toggle

Type /phi to enable or disable. Visual indicator in the status bar. All restrictions apply immediately and persist across sessions.

/phi
PHI Compliance Mode Status ENABLED Provider ollama (local only) Audit ~/.emprise/phi-audit.log Restrictions active: ✗ clipboard_read / clipboard_write ✗ web_search / web_fetch ✗ ssh_run ✓ Local file tools only ✓ All queries stay on-device

Developer Environment Audit

/dev scans 40+ tools across multiple categories. Missing something? Select it and emprise installs it with platform-aware package managers.

/dev > Languages
Developer Environment > Languages ✓ nvm ★ ~/.nvm ✓ Node.js ★ v22.1.0 ✓ Go ★ go1.23.0 ✗ Rust not found curl ... | sh ✓ Python ★ 3.12.0 > ✗ Java not found brew install openjdk ← Back
/dev > Summary
Developer Environment Summary ✓ 32 tools installed ✗ 8 tools missing Categories: ✓ Languages 4/6 ✓ Package Mgrs 3/3 ✓ Containers 2/2 ! Editors 1/3 ✓ Version Control 3/3 ! Cloud CLIs 0/4 Select any missing tool to install

Download

Fastest:

brew install senzalldev/tap/emprise                 # CLI
brew install --cask senzalldev/tap/emprise-desktop  # Desktop (macOS)
curl -fsSL https://emprise.dev/install.sh | sh

CLI and Desktop app available for macOS, Linux, and Windows.

PlatformDesktopCLINotes
macOS (Apple Silicon) Desktop DMG CLI Both signed + notarized — no Gatekeeper warnings
Windows (amd64) NSIS Installer CLI .exe Bootstraps WebView2 if missing
Linux (amd64) tar.gz CLI Needs libwebkit2gtk-4.0-37 + libgtk-3-0
$ curl -fsSL https://raw.githubusercontent.com/senzalldev/emprise-app/main/install.sh | sh

Install script auto-detects your platform on macOS and Linux. Update anytime with emprise update — checks on launch and atomically swaps in the new binary.

Source is private. If a binary you need isn't in the latest release, please file an issue — it's a CI matter, not a build-it-yourself one.

Safety

Blocked Commands

rm -rf /, dd, mkfs and other destructive operations are hard-blocked and will never run.

Protected Paths

/etc, /usr, /System, and home dotfiles are protected from modification.

Inline Visibility

Every tool call is shown with its arguments and output. Nothing runs hidden.

Async + Cancellable

Tools run asynchronously. Press esc or ctrl+c to cancel any operation at any time.

No Exfiltration

Your data is only sent to the LLM provider you configured. Nothing else phones home.

338 Tests

Comprehensive test suite covering tools, prompts, config, providers, and safety mechanisms.

Community

GitHub

Source, releases, wiki

Discussions

Questions, ideas, show and tell

Issues

Bug reports, feature requests

Wiki

Guides, commands, configuration