Atlas / MCP servers / aurite-ai / kai

kaiBLOCK

Name: kai
Author: aurite-ai

mcp/aurite-ai/kai

MCP server by aurite-ai · ai coding assistant · ai memory · claude code · Audited 2026-10-10

🧠 Kai is a context engineering platform. It's persistent memory for AI coding copilots. Teach it once, surface context automatically. MCP server for Claude Code.

Verdict

BLOCK

Grade

Trust score

57 /100

Exposed tools

169 128r · 40w · 1d

Transport

stdio

License

—

Stars

Overview

From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.

🧠 Kai Your AI copilot's memory. Persistent context across sessions, projects, and teams. Give your coding agent the context it needs — automatically.

Works with Claude Code · more copilots coming soon

The Problem

Every time you start a new conversation with your AI copilot, it forgets everything.

🔄 You repeat the same context about your project, your team, your standards
🤷 The copilot makes mistakes you've already corrected in past sessions
📄 Your policies, specs, and business rules sit in files the copilot never sees
🧠 Decisions and rationale from past conversations are lost forever

Copilots are powerful — but they have amnesia.

The Solution

Kai gives your copilot a persistent memory that grows smarter over time.

How it works: Kai runs as an MCP server alongside your copilot. You teach it your context once — policies, specs, decisions, p

Read from source at commit e17329867ef8OBSERVED · 2026-10-10

Connect

Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.

claude-code

claude mcp add kai-testing --env ANTHROPIC_API_KEY=${ANTHROPIC_API_KEY} --env KAI_USAGE_API_KEY=${KAI_USAGE_API_KEY} --env OPENAI_API_KEY=${OPENAI_API_KEY} --env PERPLEXITY_API_KEY=${PERPLEXITY_API_KEY} -- npx -y @aurite-ai/[email protected]

claude-desktop

{
  "mcpServers": {
    "kai-testing": {
      "command": "npx",
      "args": [
        "-y",
        "@aurite-ai/[email protected]"
      ],
      "env": {
        "ANTHROPIC_API_KEY": "${ANTHROPIC_API_KEY}",
        "KAI_USAGE_API_KEY": "${KAI_USAGE_API_KEY}",
        "OPENAI_API_KEY": "${OPENAI_API_KEY}",
        "PERPLEXITY_API_KEY": "${PERPLEXITY_API_KEY}"
      }
    }
  }
}

Exposed tools (169)

128 read · 40 write · 1 destructive. Blast radius: 1 tool can delete or overwrite — an agent that can be talked into calling a tool can be talked into calling this one.

Tool	Risk	Description
`LangGraph`	read	Python agent framework using LangGraph for stateful workflows
`amount`	read	Amount in cents
`append-row`	read	Append a row to a spreadsheet
`avatar`	read	Avatar URL or base64 image
`backup`	write	Create a database backup
`blocks`	read	Block Kit blocks
`body`	read	Email body (HTML or plain text)
`bucket`	read	Bucket name
`call-function`	read	Call a Supabase Edge Function
`cancel_url`	read	URL to redirect if cancelled
`capture-payment`	read	Capture an approved payment
`capture_id`	read	Capture ID to refund
`categorize_file`	read	Categorize file
`channel`	read	Channel ID or name
`channel_id`	read	Channel ID
`chat-completion`	read	Generate a chat completion
`collection`	read	Collection name
`content`	read	Email content
`count`	read	Number of connections to return
`count-tokens`	read	Count tokens in text
`create-channel`	write	Create a new channel in a team
`create-contact`	write	Create a new contact
`create-folder`	write	Create a folder in Google Drive
`create-issue`	write	Create a new issue in a repository
`create-page`	write	Create a new page in Notion
`create-payment`	write	Create a new payment order
`create-payment-intent`	write	Create a payment intent
`create-record`	write	Create a new record
`create-webhook`	write	Create a webhook for a channel
`currency`	read	Currency code
`data`	read	Record data
`database_id`	read	Database ID
`description`	read	Channel description
`display_name`	read	Channel name
`doc_id`	read	Document ID
`documents`	write	Documents to insert
`download-file`	read	Download a file from Google Drive
`email`	read	Contact email
`embed-content`	read	Generate embeddings for text
`embeddings`	read	Generate embeddings for text
`embeds`	read	Message embeds
`end_date`	read	End date (ISO format)
`execute`	destructive	Execute a SQL statement (insert, update, delete)
`fetch_api_documentation`	read	Fetch and extract relevant information from an API documentation URL. Returns structured information about endpoints, authentication, etc.
`file_id`	read	File ID
`filter`	read	Query filter
`find`	read	Find documents in a collection
`folder_id`	read	Folder ID (root if not specified)
`from`	read	Sender email
`function_name`	read	Function name
`generate-content`	read	Generate content using Gemini
`generate_connector_manifest`	read	Generate a complete ConnectorManifest from the gathered integration information. Call this after you have collected enough information about the service.
`get-connections`	read	Get user connections
`get-contact`	read	Get a contact by email or ID
`get-doc`	read	Get a document from Firestore
`get-issue`	read	Get details of a Jira issue
`get-object`	read	Get an object from S3
`get-profile`	read	Get the authenticated user profile
`get-pull-requests`	read	List pull requests for a repository
`get-transactions`	read	Get transaction history
`get-user`	read	Get user profile by username
`get-values`	read	Get values from a spreadsheet range
`guild_id`	read	Discord guild/server ID
`health_check`	read	Check if the Kai MCP server is running correctly. This tool verifies the MCP connection is working. Actions: - ping: Confirm MCP server is alive
`id`	read	Record ID
`input`	read	Input param
`insert`	write	Insert documents into a collection
`issue_key`	read	Issue key (e.g., PROJ-123)
`issue_type`	read	Issue type (Bug, Task, Story)
`jql`	read	JQL query string
`kai_ask`	read	Quick Q&A using the project knowledge base. USE THIS TOOL WHEN: - Mid-task and need specific information - User asks a direct question about the project - Need clarification on a decision or pattern - Need information beyond what
`kai_discover_integration`	write	Discover and add a new integration connector for any service, API, or tool. Use this when: - The user needs an integration that isn
`kai_initialize`	write	Deploy agent-dev rules and optionally run onboarding to collect org/user context. USE THIS TOOL WHEN: - User says
`kai_learn`	write	Send files or folders to Kai to learn from and add to the knowledge base. USE THIS TOOL WHEN: - User shares files/folders and wants Kai to
`kai_prepare_context`	read	Prepare the .kai/context-guide.md with relevant knowledge for a task. USE THIS TOOL WHEN: - Starting any new task or feature - User describes what they want to build - Before beginning implementation work - User asks
`kai_usage`	read	View token usage and cost summary for the current project. USE THIS TOOL WHEN: - User asks about usage, costs, or token consumption - User wants to see how much they
`kai_use_integration`	write	Execute an operation on a discovered integration. Use this tool to interact with external services (APIs, databases, etc.) that Kai has learned about. The integration must first be discovered using kai_learn on files that describe it. Examples: - Query a database: integration=
`kai_verify_integration`	read	Verify that an integration is correctly configured and can connect. Use this tool to test if an integration is working before using it. Checks that credentials are available and attempts a simple operation to verify connectivity. Can verify a single integration or all discovered integrations.
`key`	read	Object key
`labels`	write	Labels to apply
`limit`	read	Max results
`list-channels`	read	List available Slack channels
`list-customers`	read	List customers
`list-files`	read	List files in Google Drive
`list-messages`	read	List emails from inbox
`list-objects`	read	List objects in a bucket
`list-repos`	read	List repositories for the authenticated user or organization
`list-teams`	read	List teams the user is a member of
`list_existing_connectors`	read	List connectors that are already registered. Use this to check if a connector already exists before creating a new one.
`list_knowledge_files`	read	List all files in the knowledge base with summaries and metadata. Use this to decide which files to read.
`maxResults`	read	Maximum number of results
`max_results`	read	Maximum results (10-100)
`max_tokens`	read	Maximum tokens
`merge`	write	Merge with existing data
`messages`	read	Chat messages
`mime_type`	read	MIME type
`model`	read	Model to use
`name`	read	File name
`objectType`	read	Salesforce object type
`op-1`	read	First operation
`op-2`	read	Second operation
`order_by`	write	Field to order by
`order_id`	write	PayPal order ID
`org`	read	Organization name (optional)
`output_path`	write	Path to save backup file
`owner`	read	Repository owner
`page_id`	read	Page ID
`page_size`	read	Number of results per page
`params`	read	Query parameters
`parent_id`	read	Parent folder ID
`parse_openapi_spec`	read	Parse an OpenAPI/Swagger specification and extract operations, authentication methods, and schemas. Returns structured connector information.
`post-share`	write	Share content on LinkedIn
`post-tweet`	write	Post a new tweet
`prefix`	read	Key prefix
`priority`	read	Priority level
`project_key`	read	Project key (e.g., PROJ)
`prompt`	read	Input prompt
`properties`	read	Contact properties
`put-object`	write	Upload an object to S3
`query`	write	Run a query
`query-collection`	read	Query documents from a collection
`query-database`	read	Query a Notion database
`range`	read	A1 notation range (e.g., Sheet1!A1:D10)
`read_knowledge_file`	read	Read the full content of a specific knowledge base file. Use after listing files to read ones relevant to the task/question.
`refund`	read	Refund a captured payment
`register_connector`	read	Register the generated connector manifest in the local registry. This makes the connector available for use.
`reply_to`	read	Tweet ID to reply to
`repo`	read	Repository name
`report_contradictions`	read	Report files in the knowledge base that contradict the new file being categorized. Use the full path (subdirectory/slug) from the list_knowledge_files output.
`return_url`	read	URL to redirect after approval
`search-issues`	read	Search issues using JQL
`search-tweets`	read	Search for tweets
`search_integration_info`	read	Search for API documentation, authentication methods, and endpoint information for a service. Returns relevant documentation snippets.
`select`	read	Columns to select
`select_files_for_context`	read	Select files
`select_framework`	read	Select a framework scaffold when the task involves building an agent, workflow, or LLM-powered application
`send-email`	write	Send an email via Gmail
`send-message`	write	Send a message to a Slack channel
`send-webhook`	write	Send a message via webhook
`set-doc`	write	Set/create a document in Firestore
`soql`	read	SOQL query
`sorts`	read	Sort conditions
`spreadsheet_id`	read	Spreadsheet ID
`sql`	write	SQL query to execute
`start`	write	Pagination start index
`start_date`	write	Start date (ISO format)
`state`	read	PR state (open, closed, all)
`subject`	read	Email subject
`summary`	read	Issue summary
`table`	read	Table name
`tables`	read	Tables to backup (all if not specified)
`team_id`	read	Team ID
`test-operation`	read	A test operation
`text`	read	Message text
`title`	read	Issue title
`to`	read	Recipient email address
`transition-issue`	write	Move issue to a different status
`transition_id`	read	Transition ID
`type`	read	Repository type (all, public, private)
`update`	write	Update data in a table
`update-page`	write	Update an existing Notion page
`update-record`	write	Update an existing record
`update-values`	write	Update values in a spreadsheet range
`upload-file`	write	Upload a file to Google Drive
`username`	read	Twitter username
`values`	write	Values to write (2D array)
`visibility`	write	Post visibility (PUBLIC, CONNECTIONS)
`webhook_url`	read	Webhook URL
`where`	read	Where conditions

Trust audit

BLOCKgrade F · trust 57/100 Do not install this without reading the findings. The audit found something that could harm you or your machine.

Layer	What it checks	Result
L0	Provenance & inventory	PASS
L1	Static analysis of the code	FAIL
L2	Instruction surface (what it tells the agent)	PASS
L3	Class-specific surface	WARN
L4	Behavioural (sandbox)	SKIPPED

What the source does

Filesystem: declared (3 observation(s))
Network: declared (3 observation(s))
Shell: declared (5 observation(s))
Dependencies: not all pinned
Secrets in source: found