Inistate MCP Server
Enables module discovery, entry management, and activity submission on the Inistate platform.
README
Inistate MCP Server
MCP server for the Inistate platform — module discovery, entry management, and activity submission.
Setup
Environment Variables
| Variable | Required | Default | Description |
|---|---|---|---|
INISTATE_API_TOKEN |
Yes | — | Bearer token for Inistate API authentication |
INISTATE_API_BASE |
No | https://api.inistate.com |
API base URL |
INISTATE_MCP_MODE |
No | configure |
Initial mode: runtime, configure, or frontend (see Modes) |
INISTATE_MCP_NO_SETUP |
No | — | Set to 1 to force server mode from a terminal (skip the interactive wizard) |
INISTATE_DEBUG_FILE |
No | — | Set to 1 to log write-path tool calls to ./debug.log, or to a path to log there. Off by default; logs identifiers only, never field values |
Install from npm (recommended)
No clone or build needed — npx will fetch and run the published package on demand:
npx -y inistate-mcp
Or install globally:
npm install -g inistate-mcp
inistate-mcp
Interactive setup (recommended)
Run the binary in a terminal with no MCP client attached and it walks you through entering your API token and picks the right config file for your client:
npx -y inistate-mcp
# or, explicitly:
npx -y inistate-mcp setup
Supported clients: Claude Desktop, Claude Code (global or project-local .mcp.json), Cursor, Windsurf, Codex CLI, VS Code (user profile or workspace .vscode/mcp.json), Cline, Gemini CLI (global or workspace). Pick "Print config only" to get a JSON block to paste anywhere else.
The wizard only runs when stdin is a TTY (i.e., you launched it yourself). When an MCP client spawns the binary via piped stdio, it skips the wizard and runs as a normal MCP server — set INISTATE_MCP_NO_SETUP=1 if you need to force server mode from a terminal.
Claude Desktop Configuration
Add to your claude_desktop_config.json:
{
"mcpServers": {
"inistate": {
"command": "npx",
"args": ["-y", "inistate-mcp"],
"env": {
"INISTATE_API_TOKEN": "your-token-here"
}
}
}
}
Claude Code Configuration
claude mcp add inistate -e INISTATE_API_TOKEN=your-token-here -- npx -y inistate-mcp
Install from source
git clone https://github.com/Inistate/inistate-mcp.git
cd inistate-mcp
npm install
npm run build
Then point your MCP client at node /absolute/path/to/inistate-mcp/build/index.js.
Tools
Tools marked (configure) are only exposed in configure mode — see Modes. Tools the active backend cannot serve (e.g. scaffold_module on the hosted Platform) stay registered but return a structured capability message instead of failing silently.
| Tool | Description |
|---|---|
list_workspaces |
List workspaces the user has access to |
set_workspace |
Set the active workspace |
list_modules |
List all discoverable modules in the workspace |
get_module_schema |
Get the canvas schema (basic or extended tier) — available in every mode |
get_module_canvas |
Get full module definition with stable IDs (round-trippable) (configure) |
list_entries |
Query entries with filters, sorting, and pagination |
get_entry |
Read a single entry by ID |
get_form |
Get form fields and defaults for an activity |
submit_activity |
Create, edit, delete, or run custom activities |
submit_activities |
Bulk variant — same activity applied to up to 100 entries in one call |
get_entry_history |
Get entry audit trail and comments |
request_upload_url |
Default upload path — get a presigned S3 URL to PUT file bytes to |
confirm_upload |
Confirm a presigned upload completed; returns the File/Image field path |
upload_file |
Fallback upload via base64/multipart (use only if the presigned flow fails) |
download_file |
Download a file (returns pre-signed URL) |
design_workflow |
Generate a scaffolded module template from a description (configure) |
validate_design |
Validate a module schema before creating or updating (configure) |
create_module |
Create a new module with schema (configure) |
update_module |
Update an existing module's schema (configure) |
scaffold_module |
Draft a module schema from existing data (SQLite, Notion, or Airtable table) (configure) — served by the local runtime (inistate-core); on the hosted Platform backend it returns a capability message pointing to design_workflow |
switch_mode |
Switch the active mode (runtime / configure / frontend) |
Resources
| URI | Description |
|---|---|
inistate://modules |
List all modules |
inistate://modules/{name}/canvas |
Basic module schema (fields + states) |
inistate://modules/{name}/canvas/extended |
Extended schema with activities and flows |
inistate://guardrails |
Server-enforced submit_activity rules (read once per session) |
inistate://schema/runtime |
Runtime schema — entry/activity/file types and filter operators (default) |
inistate://schema/configure |
Module-design schema — write format, field types, colors (configure) |
inistate://design-guide |
FACTS Module Design Guide (configure) |
inistate://frontend-guide |
REST API reference for hand-written UIs (frontend) |
Prompts
| Prompt | Description |
|---|---|
design_factsops_workflow |
Guide an agent through designing a complete workflow module (configure) |
execute_activity |
Guide an agent through executing a specific activity |
diagnose_entry |
Guide an agent through investigating an entry's state and history |
modify_module |
Guide an agent through modifying an existing module's schema (configure) |
Modes
The server exposes a focused tool/resource surface depending on the active mode, keeping agent context lean. Use switch_mode to change it, or set the initial mode via the INISTATE_MCP_MODE env var (default: configure).
| Mode | Surface |
|---|---|
runtime |
Entry and activity operations only — querying, reading, submitting, files, history. The leanest surface for using existing modules. |
configure |
Everything in runtime plus the module-design tools, resources, and prompts (marked (configure) above). |
frontend |
Everything in configure plus the inistate://frontend-guide resource for building hand-written UIs against the REST API. |
Tools and resources marked (configure) / (frontend) are absent from the tool list in narrower modes — switch modes to reveal them.
Typical Workflow
list_workspaces→set_workspace— select a workspace (auto-selected when exactly one matches; both return the workspace's module list, solist_modulesis only needed to refresh)get_module_schema— understand a module's fields, states, and activitiesget_form— discover required fields before the first submission per (module, activity); reuse its schema for further entriessubmit_activity— create or update entries (submit_activitiesfor bulk)list_entries— query and browse data (use thefieldsparameter to keep payloads small)get_entry_history— review entry history
Development
npm run watch # Watch mode for TypeScript compilation
npm run inspector # Test with MCP Inspector
MCP Setup
- Setup
curl -L "https://github.com/modelcontextprotocol/registry/releases/latest/download/mcp-publisher_$(uname -s | tr '[:upper:]' '[:lower:]')_$(uname -m | sed 's/x86_64/amd64/;s/aarch64/arm64/').tar.gz" | tar xz mcp-publisher && sudo mv mcp-publisher /usr/local/bin/
or
$arch = if ([System.Runtime.InteropServices.RuntimeInformation]::ProcessArchitecture -eq "Arm64") { "arm64" } else { "amd64" }; Invoke-WebRequest -Uri "https://github.com/modelcontextprotocol/registry/releases/latest/download/mcp-publisher_windows_$arch.tar.gz" -OutFile "mcp-publisher.tar.gz"; tar xf mcp-publisher.tar.gz mcp-publisher.exe; rm mcp-publisher.tar.gz
- Verify
mcp-publisher --help
- Authenticate
mcp-publisher login github
- Publish: see below
Packaging & Versioning
# Example adding new feature
git checkout -b feat/add-user-tool
# After coding
npx changeset
# Choose:
#
# minor
# Added new user search tool
# Release
npm run release
# This does:
# install dependencies
# test
# bump version + update changelog + sync server.json
# validate MCP server config
# build (via npm prepare hook)
# publish to npm
# publish to MCP registry
PM2 (Ubuntu/AWS)
Run the HTTP transport in production using PM2:
npm install
npm run build
npm run pm2:start
npx pm2 save
Enable startup on reboot:
sudo npx pm2 startup systemd -u ubuntu --hp /home/ubuntu
npx pm2 save
Common operations:
npm run pm2:restart
npm run pm2:logs
npm run pm2:stop
Set required environment variables (INISTATE_API_TOKEN, and optionally INISTATE_API_BASE, INISTATE_WORKSPACE_ID, OAUTH_ISSUER_URL, INISTATE_APP_URL) in your shell, PM2 ecosystem env, or deployment secret manager before starting.
Testing
Run all tests
npm test
Watch mode (re-runs on file changes)
npm run test:watch
Test structure
Tests are in src/ alongside the source files and use Vitest:
| File | Type | What it covers |
|---|---|---|
src/schema.test.ts |
Unit tests (76) | designWorkflow, validateDesign (including platform parity and input normalization), helper functions (isValidFieldType, isValidColor, isValidActor, suggestColorForState) |
src/activity-guard.test.ts |
Unit tests (42) | submit_activity guard rules — human/hybrid actor, state-change confirmation, confidence-inflation, reference-shape validation |
src/tools.schema.test.ts |
Unit tests (19) | Tool input-schema shapes and validation |
src/backend-capabilities.test.ts |
Unit tests (9) | Capability gating — tools the active backend cannot serve return a capability message |
src/flagged-annotation.test.ts |
Integration tests (5) | Flagged-response annotation — suppressed transitions are explained (flag_reason + agent_action) so agents stop retrying with higher confidence |
src/server.test.ts |
Integration tests (17) | Spins up the MCP server as a child process and exercises it through the official MCP SDK client — mode-gated tool/resource/prompt discovery, switch_mode, resource reads, prompt retrieval, and local tool calls |
Unit tests cover:
- Field type and color validation against the schema
- State color suggestion logic
- Design validation: duplicate names, invalid types/colors/actors, initial state rules, flow integrity, unreachable states, unused activities, AI confidence warnings
- Input normalization: field-type, state-color, and industry aliases; parsing states from a description
- Workflow design: pattern detection (approval, ticket, pipeline, record list), industry defaults
Integration tests verify (no API token needed):
- Mode-gated tool/resource/prompt discovery — runtime mode hides the configure surface,
switch_modereveals and collapses it design_workflow,validate_designwork end-to-end through the MCP protocol- Static resources (
inistate://schema/runtime,inistate://design-guide) return valid content - All 4 prompts return correctly templated messages
Interactive testing with MCP Inspector
INISTATE_API_TOKEN=your-token npm run inspector
Opens a browser UI where you can interactively call tools, inspect schemas, and see responses.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.