ChatGipite by Lugitech

"Ang Chat bot ng mga Gipit"

Business idea generation, validation, and execution planning as an MCP server. Runs a full pipeline from raw idea to financial model, pitch deck, and go-to-market playbook — all stored locally per idea.

Prerequisite: Node.js 20+ and npm install (or npx chatgipite for one-shot use without a local clone).

Claude Code: install as a skill (recommended)

Skills load on demand — ~50 tokens of overhead until invoked, vs ~5–8K for the full MCP server. For Claude Code users, this is the lightest option.

./install.sh -c skill        # macOS / Linux / Git Bash
install.bat -c skill         REM Windows (admin shell needed for junction)

Installs to ~/.claude/skills/chatgipite/ as a symlink/junction back to this repo. Updating the repo updates the skill in place. Inside Claude Code, just ask: "validate this idea: …" — the skill is auto-discovered and dispatches to the same handler code as the MCP server.

For non-Claude-Code clients (Cursor, Zed, Gemini CLI, Codex, etc.), keep using the MCP install paths below. Both can coexist.

Quick Start

Linux / macOS / Git Bash (Windows):

cd ChatGipite
./install.sh                              # Claude Desktop
./install.sh -c claude                    # Claude Code (workspace-local)
./install.sh -c claude --global           # Claude Code (global user config)
./install.sh -c skill                     # Claude Code skill (recommended for Claude Code)
./install.sh -c cursor                    # Cursor (workspace-local)
./install.sh -c cursor --global           # Cursor (global)
./install.sh -c windsurf                  # Windsurf (global only)
./install.sh -c vscode                    # VS Code (.vscode/mcp.json)
./install.sh -c gemini                    # Gemini CLI (workspace-local)
./install.sh -c gemini --global           # Gemini CLI (global)
./install.sh -c codex                     # OpenAI Codex CLI (workspace-local)
./install.sh -c codex --global            # OpenAI Codex CLI (global)
./install.sh -c zed                       # Zed (global)
./install.sh -c kilo                      # Kilo Code
./install.sh -c opencode                  # OpenCode (workspace-local)
./install.sh -c opencode --global         # OpenCode (global)
./install.sh -c goose                     # Goose
./install.sh -c all                       # all detected MCP clients (skill is opt-in)

Windows (Command Prompt / PowerShell):

cd ChatGipite
install.bat                               REM Claude Desktop
install.bat -c claude                     REM Claude Code (workspace-local)
install.bat -c claude --global            REM Claude Code (global user config)
install.bat -c skill                      REM Claude Code skill (recommended for Claude Code)
install.bat -c cursor                     REM Cursor (workspace-local)
install.bat -c cursor --global            REM Cursor (global)
install.bat -c windsurf                   REM Windsurf (global only)
install.bat -c vscode                     REM VS Code (.vscode/mcp.json)
install.bat -c gemini                     REM Gemini CLI (workspace-local)
install.bat -c gemini --global            REM Gemini CLI (global)
install.bat -c codex                      REM OpenAI Codex CLI (workspace-local)
install.bat -c codex --global             REM OpenAI Codex CLI (global)
install.bat -c zed                        REM Zed (global)
install.bat -c kilo                       REM Kilo Code
install.bat -c opencode                   REM OpenCode (workspace-local)
install.bat -c opencode --global          REM OpenCode (global)
install.bat -c goose                      REM Goose
install.bat -c all                        REM all detected MCP clients (skill is opt-in)

The installer runs npm install, writes the MCP config entry, and validates the server.

Supported MCP Clients

Client	`-c TYPE`	Config written	Notes
Claude Desktop	`claudedesktop`	OS-specific `claude_desktop_config.json`	Restart required
Claude Code	`claude`	`.mcp.json` (workspace) or `~/.claude.json` (global)	Use `--global` for user scope
Cursor	`cursor`	`.cursor/mcp.json` or `~/.cursor/mcp.json` (global)	Use `--global` for global
Windsurf	`windsurf`	`~/.codeium/windsurf/mcp_config.json`	Global only
VS Code	`vscode`	`.vscode/mcp.json`	Workspace-local; global via VS Code settings UI
Gemini CLI	`gemini`	`.gemini/settings.json` or `~/.gemini/settings.json` (global)	Use `--global` for global
Codex CLI	`codex`	`.codex/config.toml` or `~/.codex/config.toml` (global)	TOML; use `--global` for global
Zed	`zed`	`~/.config/zed/settings.json`	Global only
Kilo Code	`kilo`	`.kilocode/mcp.json`	Workspace-local only
OpenCode	`opencode`	`opencode.json` / `~/.config/opencode/opencode.json`	Use `--global` for global
Goose	`goose`	`~/.config/goose/config.yaml`	Global only
pi.dev	`pidev`	n/a	Prints manual instructions; no auto-config
Claude Code skill	`skill`	`~/.claude/skills/chatgipite/` (symlink)	Lowest context cost; Claude Code only
All above	`all`	All detected existing configs	Skips clients not yet installed

Installer Flags

  -c, --client TYPE   claudedesktop, claude, cursor, windsurf, vscode, gemini, codex,
                      zed, kilo, opencode, goose, pidev, skill, all  (default: claudedesktop)
  -f, --force         Skip prompts, overwrite existing config
  -u, --uninstall     Remove from MCP client config
      --upgrade       Upgrade deps and reconfigure (alias: --update)
      --status        Show where this server is currently installed
      --global        Write to global config (claude, cursor, gemini, codex, opencode)
      --skip-test     Skip server validation
  -h, --help          Show this help

Check install status

./install.sh --status

Upgrade

Pull the latest source first (or re-download and extract), then:

./install.sh --upgrade                    # reinstall deps
./install.sh --upgrade -c all             # also reconfigure all clients

--update is an alias for --upgrade.

Uninstall

./install.sh -u -c all
install.bat -u -c all   # Windows

Manual Setup

npm install
node server.js   # verify it starts, then Ctrl+C

No API key required for fully local use — set default_provider: ollama in config/providers.yaml. For Anthropic/Claude: export ANTHROPIC_API_KEY=sk-ant-...

Add ChatGipite to your MCP client config (use absolute paths).

Most clients use the same mcpServers JSON — add this block to the config file for your client:

{
  "mcpServers": {
    "chatgipite": {
      "command": "node",
      "args": ["/absolute/path/to/ChatGipite/server.js"]
    }
  }
}

Client	Config file
Claude Desktop	`%APPDATA%\Claude\claude_desktop_config.json` (Win) · `~/Library/Application Support/Claude/claude_desktop_config.json` (macOS) · `~/.config/Claude/claude_desktop_config.json` (Linux)
Claude Code	`.mcp.json` (workspace) or `~/.claude.json` (global)
Cursor	`.cursor/mcp.json` (workspace) or `~/.cursor/mcp.json` (global)
Windsurf	`~/.codeium/windsurf/mcp_config.json`
Gemini CLI	`.gemini/settings.json` (workspace) or `~/.gemini/settings.json` (global)
Kilo Code	`.kilocode/mcp.json`

Claude Code global (CLI):

claude mcp add --scope user chatgipite -- node /absolute/path/to/ChatGipite/server.js

Clients with different config format — use the same command/args values, different structure:

Client	Config file	Key difference
VS Code	`.vscode/mcp.json`	Top-level key is `servers` (not `mcpServers`), add `"type": "stdio"`
Zed	`~/.config/zed/settings.json`	Top-level key is `context_servers`, command is nested: `{ "path": ..., "args": ..., "env": {} }`
OpenAI Codex	`.codex/config.toml`	TOML format: `[mcp_servers.chatgipite]`, `command = "node /absolute/path/to/ChatGipite/server.js"`
OpenCode	`opencode.json` or `~/.config/opencode/opencode.json`	Top-level key is `mcp` (not `mcpServers`)
Goose	`~/.config/goose/config.yaml`	YAML format: under `extensions`, uses `cmd`/`args`/`type: stdio`/`enabled: true`

On Windows, use C:\absolute\path\to\ChatGipite\server.js. Restart the client after editing any config.

Tools

Idea generation & validation

Tool	Output
`biz_generate`	Business ideas from a sector, problem, or constraints (`count` param, default 1)
`biz_validate`	Viability analysis: problem, ICP, solution, revenue model, market size, ICE score
`biz_ice_score`	Impact / Confidence / Ease scoring (1-10 each) with rationale

Market & customer analysis

Tool	Output
`biz_tam`	Deep TAM/SAM/SOM sizing — top-down + bottom-up with 5-year projections
`biz_personas`	3-5 JTBD-framed customer personas, ICP priority matrix, WTP per segment
`biz_trends`	PESTLE + technology S-curve analysis, tailwinds/headwinds, market timing verdict
`biz_competitors`	Competitor table, gap analysis, differentiation strategy
`biz_landscape`	Deep competitive intelligence: funding trajectories, moat analysis, white-space map, positioning grid

Strategy & positioning

Tool	Output
`biz_swot`	SWOT table + Porter's Five Forces (rated 1-5) with SO/WO/ST/WT cross-analysis
`biz_pricing`	3-tier pricing architecture, value metric, competitor price table, expansion revenue paths
`biz_gtm`	GTM motion, beachhead segment, 0→100 and 100→1000 channels, 30-day launch plan
`biz_journey`	7-stage customer journey map, moments of truth, drop-off risks, delight opportunities
`biz_expansion`	Market entry modes, 3-phase expansion roadmap, localization requirements, go/no-go criteria

Financial modeling & risk

Tool	Output
`biz_financials`	Unit economics, LTV:CAC ratio, break-even, 12-month P&L
`biz_model`	3-scenario P&L (base/upside/downside), cohort retention, sensitivity table, funding requirements
`biz_risks`	Risk register (probability × impact), scenario narratives, regulatory flags, 90-day assumption tests

Presentation & execution

Tool	Output
`biz_canvas`	Strategy canvas; `canvas_type`: bmc (default) / lean / vpc / mission / ai-platform
`biz_lean_canvas`	Lean Canvas (Running Lean 3rd ed) — best for unproven ideas
`biz_value_prop`	Value Proposition Canvas (jobs/pains/gains ↔ value map)
`biz_mission_canvas`	Mission Model Canvas (mission-driven / non-profit / gov)
`biz_ai_canvas`	AI Canvas + Platform Canvas (model-core or multi-sided)
`biz_pitchdeck`	10-slide pitch deck content
`biz_playbook`	30/60/90-day execution plan with milestones, KPIs, risk flags
`biz_name`	Name candidates with live domain and social availability check
`biz_name_check`	Availability check for a name: domains via RDAP→WHOIS→DNS (com/io/ai/co/app/dev/net/org) + 9 socials (X, Instagram, TikTok, LinkedIn, GitHub, YouTube, Facebook, Reddit, Threads). Also a standalone CLI: `node scripts/check-name.mjs "Name"`. Optional: `CHATGIPITE_WMN=1` uses the maintained WhatsMyName signatures (fixes TikTok etc.; `npm run update-wmn` to refresh). `CHATGIPITE_PLAYWRIGHT=1` + `npx playwright install chromium` adds a headless-browser fallback. IG/X stay "verify manually" (login-walled — nothing logged-out resolves them).
`biz_synthesis`	Executive brief: conviction score (1-10), 5 key decisions, 90-day action roadmap
`biz_north_star`	North Star Metric + input model + Customer Factory + guardrails
`biz_rice_score`	RICE prioritization (Reach × Impact × Confidence / Effort)
`biz_assumptions`	Riskiest-assumption map + Test Card (hypothesis/metric/threshold/timebox)
`biz_prfaq`	Working-Backwards press release + internal/customer FAQ
`biz_mvp`	Smallest MVP scope to test the riskiest assumption

Perspectives & psychology

These accept either idea_slug (a saved idea) or product (raw text for a new/existing product).

Tool	Output
`biz_perspectives`	6 independent blind lenses (contrarian, customer, operator, investor, regulator, futurist) run in parallel + a reviewer reconciling agreements / conflicts / blind-spots into a net verdict
`biz_product_psych`	Product/UX behavioral design (Fogg B=MAP, Hook, friction, peak-end) + technique-density audit (under-leveraged / balanced / over-engineered) + dark-pattern flags. `mode`: apply / audit / full
`biz_marketing_psych`	Persuasion / conversion psychology (Cialdini 7, loss aversion, anchoring, decoy) + technique-density audit + dark-pattern flags. `mode`: apply / audit / full
`biz_engagement`	App-addiction specialist: detects abusive/addictive patterns (infinite scroll, slot-machine rewards, streak guilt, manipulative notifications, fake urgency, roach-motel cancel) with severity + regulatory exposure (FTC / EU DSA Art.25 / CPPA), runs regret/transparency/exit ethics tests, then prescribes humane engagement + retention (rewrite playbook). Grounded in a cited knowledge base. `mode`: audit / apply / full

Pipelines

Tool	Output
`biz_full_run`	7-step pipeline: validate → competitors → financials → canvas → pitchdeck → playbook → names
`biz_deep_run`	24-step workflow + names: functional analysis (validate → personas → trends → market → TAM → financials → model → SWOT → pricing → GTM → journey → risks → landscape → expansion) + 6 perspective lenses → panel review + product/marketing psychology → synthesis → names
`biz_incubate`	Incubation loop: validate → assumptions → lean canvas + value-prop + north star → MVP → pivot/persevere decision
`biz_launch_sprint`	Working-Backwards launch: validate → PR/FAQ → MVP → GTM → launch checklist
`biz_recall`	Full-text search across all stored analyses

Typical Flow

Step by step (standard):

biz_generate    {sector: "healthtech"}
biz_validate    {idea: "AI triage assistant for rural clinics"}
biz_competitors {idea_slug: "ai-triage-assistant-for-rural-clinics"}
biz_financials  {idea_slug: "ai-triage-assistant-for-rural-clinics"}
biz_canvas      {idea_slug: "ai-triage-assistant-for-rural-clinics"}
biz_pitchdeck   {idea_slug: "ai-triage-assistant-for-rural-clinics"}
biz_name        {idea_slug: "ai-triage-assistant-for-rural-clinics"}
biz_playbook    {idea_slug: "ai-triage-assistant-for-rural-clinics"}

Full pipeline in one call:

biz_full_run      {idea: "AI triage assistant for rural clinics"}
biz_deep_run      {idea: "AI triage assistant for rural clinics"}   ← deep version (24 steps, ~25 artifacts)
biz_incubate      {idea: "AI triage assistant for rural clinics"}   ← lean incubation loop (7 artifacts)
biz_launch_sprint {idea: "AI triage assistant for rural clinics"}   ← Working-Backwards launch (5 artifacts)

Artifacts

Saved to ideas/{slug}/:

Standard (`biz_full_run`)

File	Contents
`brief.md`	Validation analysis and ICE score
`competitive.md`	Competitor landscape
`financials.md`	Financial model
`canvas.md`	Business Model Canvas
`pitchdeck.md`	Pitch deck
`playbook.md`	30/60/90-day plan
`names.md`	Name candidates with availability table

Deep (`biz_deep_run`)

All of the above, plus:

File	Contents
`personas.md`	Customer personas, ICP matrix, WTP
`trends.md`	PESTLE, S-curve, market timing
`tam.md`	TAM/SAM/SOM sizing
`model.md`	3-scenario financial model
`swot.md`	SWOT + Porter's Five Forces
`pricing.md`	Pricing architecture
`gtm.md`	Go-to-market strategy
`journey.md`	Customer journey map
`risks.md`	Risk register and scenarios
`landscape.md`	Deep competitive intelligence
`expansion.md`	Market entry and expansion roadmap
`lens-*.md`	Six perspective lenses (contrarian, customer, operator, investor, regulator, futurist)
`perspective-review.md`	Reconciled panel review: agreements, conflicts, blind spots, net verdict
`product-psychology.md`	Product/UX behavioral-design audit + technique density
`marketing-psychology.md`	Persuasion/conversion audit + technique density
`synthesis.md`	Executive brief with conviction score

Configuration

The client LLM (what you chat with in your IDE) and ChatGipite's internal agents are independent. You can use ChatGipite from Cursor with GPT-4o while the analysis runs on a local Ollama model.

Config resolution order

ChatGipite looks for provider config in this order:

CHATGIPITE_CONFIG env var — absolute path to any yaml/json file
providers.yaml in the MCP server's working directory
providers.json in the MCP server's working directory
config/providers.yaml in the install directory (default)

Using a project-specific config

Drop a providers.yaml anywhere and point to it via the env var in your MCP config:

{
  "mcpServers": {
    "chatgipite": {
      "command": "node",
      "args": ["/absolute/path/to/ChatGipite/server.js"],
      "env": {
        "CHATGIPITE_CONFIG": "/path/to/my-project/providers.yaml"
      }
    }
  }
}

Or set it in your shell before launching the client:

export CHATGIPITE_CONFIG=/path/to/my-project/providers.yaml
# Windows
set CHATGIPITE_CONFIG=C:\path\to\my-project\providers.yaml

Supported providers

Provider	Type	Notes
`anthropic`	Native SDK	Default. Set `ANTHROPIC_API_KEY`.
`ollama`	Local HTTP	No key needed.
`openai`	OpenAI-compatible	Set `OPENAI_API_KEY`.
`openrouter`	OpenAI-compatible	Access 200+ models. Set `OPENROUTER_API_KEY`.
`groq`	OpenAI-compatible	Fast inference. Set `GROQ_API_KEY`.
`nvidia`	OpenAI-compatible	NVIDIA NIM. Set `NVIDIA_API_KEY`.
`google`	OpenAI-compatible	Gemini via OpenAI-compat endpoint. Set `GOOGLE_API_KEY`.
`lmstudio`	OpenAI-compatible	Local. No key needed. Default port `1234`.
`jan`	OpenAI-compatible	Local. No key needed. Default port `1337`.

Any endpoint following the /v1/chat/completions spec works with type: openai_compatible — including Azure OpenAI, Mistral, Together AI, Kilo, and others.

No API key configured for a provider? ChatGipite falls back to passthrough — the host model (the one you're chatting with) handles the analysis instead.

Sample config

default_provider: ollama

providers:
  ollama:
    default_model: llama3.2
    base_url: http://localhost:11434

  lmstudio:
    type: openai_compatible
    default_model: your-loaded-model-name
    base_url: http://localhost:1234/v1

  nvidia:
    type: openai_compatible
    default_model: qwen/qwen3.5-397b-a17b   # recommended on NVIDIA NIM; markedly better than llama-3.3-70b on structured analysis
    base_url: https://integrate.api.nvidia.com/v1
    api_key: nvapi-...   # or set NVIDIA_API_KEY env var

  openrouter:
    type: openai_compatible
    default_model: openai/gpt-4o
    base_url: https://openrouter.ai/api/v1
    # api_key: ...  (set OPENROUTER_API_KEY env var)

# Route specific agents to specific providers
agent_providers:
  ideator: lmstudio       # fast local generation
  validator: nvidia       # stronger model for validation
  financial-analyst: openrouter

# Override model per agent (stacks on top of agent_providers)
model_per_agent:
  financial-analyst: anthropic/claude-opus-4-5

Debug logging

Set CHATGIPITE_DEBUG=1 to enable verbose logging to chatgipite.log in the install directory:

"env": { "CHATGIPITE_DEBUG": "1", "CHATGIPITE_CONFIG": "/path/to/providers.yaml" }

Prompt cache threshold (Anthropic)

lib/llm-adapter.js only attaches cache_control when the system prompt or context block is ≥ 8192 chars (~2048 tokens, the Sonnet 4.6 minimum). Below that threshold the API silently no-ops the cache header — so smaller prompts won't show cache_read_input_tokens in debug logs. This is intentional, not a bug.

Troubleshooting

Windows: fetch failed / UNABLE_TO_VERIFY_LEAF_SIGNATURE against any HTTPS provider. On corporate networks with TLS-intercepting proxies (Zscaler, Netskope, ZTNA), Node's bundled CA store doesn't trust the company root. Run with system CA:

set NODE_OPTIONS=--use-system-ca

Or add it to the MCP env block:

"env": { "NODE_OPTIONS": "--use-system-ca", "ANTHROPIC_API_KEY": "..." }

This pulls in the OS trust store (which already has your corporate root).

Prompt A/B test harness

tests/run-ab.mjs + tests/judge.mjs compare current subagent prompts against a frozen pre-modernization baseline and a no-system-prompt vanilla variant. See tests/README.md. Requires ANTHROPIC_API_KEY for the judge step (~$2-3 per run).

Requirements

Node.js 20+ (Node 24 supported)
LLM API key (optional — Ollama works out of the box for fully local use)

Changelog

See CHANGELOG.md.

License

MIT

Name		Name	Last commit message	Last commit date
.claude-plugin		.claude-plugin
.github/workflows		.github/workflows
config		config
data		data
knowledge		knowledge
lib		lib
scripts		scripts
skill		skill
subagents		subagents
tests		tests
tools		tools
workflows		workflows
.gitattributes		.gitattributes
.gitignore		.gitignore
.mcp.json		.mcp.json
CHANGELOG.md		CHANGELOG.md
CONTRIBUTING.md		CONTRIBUTING.md
LICENSE		LICENSE
README.md		README.md
SECURITY.md		SECURITY.md
install.bat		install.bat
install.sh		install.sh
package-lock.json		package-lock.json
package.json		package.json
server.js		server.js

Latest commit

History

Folders and files

Repository files navigation

ChatGipite by Lugitech

Claude Code: install as a skill (recommended)

Quick Start

Supported MCP Clients

Installer Flags

Check install status

Upgrade

Uninstall

Manual Setup

Tools

Idea generation & validation

Market & customer analysis

Strategy & positioning

Financial modeling & risk

Presentation & execution

Perspectives & psychology

Pipelines

Typical Flow

Artifacts

Standard (biz_full_run)

Deep (biz_deep_run)

Configuration

Config resolution order

Using a project-specific config

Supported providers

Sample config

Debug logging

Prompt cache threshold (Anthropic)

Troubleshooting

Prompt A/B test harness

Requirements

Changelog

License

About

Topics

Resources

Contributing

Security policy

Stars

Watchers

Forks

Releases

Packages

Used by

Contributors

Languages

Standard (`biz_full_run`)

Deep (`biz_deep_run`)