AI & Agents

Self-Hosted Knowledge Base: Open-Source Tools, Costs, and Modern Alternatives

A self-hosted knowledge base gives organizations physical custody over their internal documentation and runbooks. While running open-source wikis in Docker containers eliminates vendor subscriptions, teams take on recurring infrastructure maintenance and database backups, and run any AI search or agent access on their own servers too. Intelligent workspaces unify team files, hybrid search, and agent connectivity as a hosted service, without the overhead of self-hosting.

Derek Labian 19 min read Updated
An intelligent workspace indexes documents on arrival for hybrid keyword and semantic retrieval, serving humans and AI agents from verified source files.

Why Teams Choose a Self-Hosted Knowledge Base and Where It Falls Short

A Gartner survey of 4,861 full-time employees, run from September to November 2022, found that 47% of digital workers struggle to find information or data needed to effectively perform their jobs. Internal knowledge management remains a critical operational hurdle for growing engineering organizations and technical teams. When company knowledge sits scattered across private chat messages, disparate local drives, and unindexed cloud storage folders, employees lose time attempting to identify which documents reflect current production architecture.

A self-hosted knowledge base is documentation software deployed on an organization's own on-premises servers or private cloud instances, granting full custody over underlying data. Teams choose self-hosting primarily to retain physical ownership of sensitive files, meet strict data sovereignty requirements, avoid recurring per-seat SaaS subscription fees, or run documentation in air-gapped environments without external internet connectivity.

Deploying a documentation tool with a container command creates an appealing first impression. Within fifteen minutes, an engineer can start a web server, configure an administrative account, and paste in initial onboarding guides. Over the following months, however, static wikis face a common operational obstacle: documentation drift. Because technical documentation lives in an isolated application away from day-to-day project files, specifications, and architecture decisions, pages decay as products evolve. Engineers update code in repositories and store design assets in cloud drives, but they rarely revisit the internal wiki to update stale paragraphs. The self-hosted server transforms from an active operational asset into an untrusted digital archive.

Why Engineering Teams Evaluate On-Premise Documentation

Organizations evaluate on-premise knowledge base software for several distinct operational reasons:

  • Data Custody and Privacy: Organizations handling sensitive intellectual property, source code algorithms, or regulated customer data hesitate to upload internal documentation to multi-tenant cloud SaaS providers. Self-hosting keeps all documentation records behind corporate perimeter firewalls.
  • Network Isolation and Air-Gapped Environments: Defense contractors, utility providers, and specialized research facilities often operate isolated networks disconnected from the public internet. On-premise knowledge base tools run on private local area networks without external telemetry dependencies.
  • Predictable Hardware Economics: SaaS documentation products typically bill on a per-user, per-month pricing model. For an enterprise with hundreds or thousands of occasional readers, subscription fees compound quickly. Running a self-hosted wiki on dedicated hardware or an existing virtualization cluster appears financially advantageous.
  • Custom Extensibility: Open-source knowledge bases allow engineering teams to inspect the source code, develop custom authentication modules, modify user interface themes, and configure database-level integrations.

The Growing Disconnect Between Static Wikis and Working Files

The fundamental flaw of traditional wiki software is its separation from active files. A software engineering team does not conduct work exclusively in markdown text blocks. Real project context spans diverse file formats: architecture diagrams, API specifications, customer contracts, design prototypes, data pipeline configurations, and spreadsheet models.

When a company deploys a standalone wiki, it forces employees to maintain two disconnected systems. The actual work happens in files stored across cloud storage accounts, while the documentation describing that work must be manually transcribed into wiki pages. Because nobody has time to update documentation manually, documentation drift sets in.

When a new engineer joins the team and searches the wiki for deployment instructions, they find outdated terminal commands from eight months earlier. To resolve the discrepancy, the engineer must message a senior colleague to ask where the current runbook lives. The senior colleague points them to a document stored in a shared drive or a local text file. At that point, the self-hosted knowledge base ceases to function as a reliable source of truth.

How Leading Open-Source Knowledge Base Tools Compare

Selecting an open-source self-hosted knowledge base requires evaluating software architecture, database dependencies, content hierarchy, and operational complexity. Different tools serve different organizational structures, ranging from structured book-style manuals to flexible collaborative wikis. While some platforms prioritize ease of setup with minimal external services, others demand multi-container orchestration with dedicated caching layers and relational database clusters.

Engineers evaluating these options must consider initial authoring convenience alongside long-term maintenance over years of operational growth. Factors like search indexing performance, database schema migrations, and user access permissions dictate whether a platform remains sustainable.

To make an informed evaluation, teams can compare five widely used self-hosted documentation platforms: BookStack, Wiki.js, Outline, DokuWiki, and Docmost. Four are open source; Outline is source-available.

Analysis of the Top Open-Source Knowledge Base Options

Each open-source knowledge base platform approaches content organization and server architecture with distinct design decisions:

  • BookStack: Built with PHP and the Laravel framework, BookStack organizes documentation into an opinionated hierarchy: Shelves, Books, Chapters, and Pages. This structure mirrors physical library organization, making it intuitive for non-technical team members, human resources departments, and operations staff. BookStack provides both a WYSIWYG editor and a Markdown editor, complete page revision history, granular role-based permissions, and built-in image management, along with a built-in REST API and OIDC, SAML2, and LDAP sign-in. Its fixed shelf, book, chapter, and page structure suits manual-style documentation but can feel restrictive for content that does not fit that shape. Search is keyword-based, with exact-phrase, tag, and filter syntax.
  • Wiki.js: Built on Node.js, Wiki.js is a feature-rich, developer-focused documentation engine released under the AGPL. The current major version, Wiki.js 3, requires PostgreSQL (versions 17 or 18) and Node.js 26, and needs its own dedicated hostname. Authors can write in Markdown, a visual editor, or AsciiDoc, and Live Notes let people co-edit markdown notes in real time. It offers two-way Git sync, so documentation can live in a Git repository, and built-in full-text search on PostgreSQL with no external search service to run. Wiki.js 3 also ships an optional, off-by-default MCP server that lets assistants such as Claude Code and Cursor search, read, and (with permission) edit pages. Its documentation says the wiki itself runs no AI model and does not embed content, leaving that to outside AI tools.
  • Outline: Outline delivers a fast, modern user experience comparable to Notion or Linear. Built with Node.js and React, Outline features real-time collaborative editing, slash commands, nested page collections, and a distraction-free interface. Each Outline workspace also includes a built-in MCP server. Its self-hosted deployment needs PostgreSQL and Redis, stores uploads on local disk or in S3-compatible storage, and has no built-in email and password sign-in, so it must be connected to an SSO, OIDC, or SAML provider; Outline's hosting guide says it requires dev-ops experience to run in production. Outline is released under the Business Source License 1.1, which its license text says is not an open source license: production use is allowed except offering Outline to third parties as a commercial document service, and each version moves to the Apache 2.0 license after four years.
  • DokuWiki: DokuWiki is a veteran open-source wiki built on PHP that operates entirely on a flat-file database. Every page is stored as a plain text file inside standard filesystem directories. Because it requires no relational database engine, installation is straightforward, system resource consumption is negligible, and backups require only copying a directory tree. DokuWiki uses page locks to keep two people from editing the same page at once, search runs on its own full-text index, and much of its functionality beyond the core comes from community plugins.
  • Docmost: Docmost is an emerging open-source collaborative wiki designed as an alternative to proprietary team wikis. Built with Node.js, NestJS, React, PostgreSQL, and Redis, Docmost provides real-time collaborative editing, spaces for organizing team departments, nested document trees, and revision history. Docmost is licensed under AGPL 3.0, and several features, including AI answers, MCP, the REST API, SSO, and page-level permissions, are in its paid editions.

Technical Stack and Deployment Architecture Comparison

The table below outlines the core technical specifications, backend storage dependencies, search mechanisms, and deployment complexity for each leading open-source knowledge base.

Documentation Tool Application Runtime Database Backend File Storage Engine Search Engine Mechanism Primary Operational Tradeoff
BookStack PHP 8.2+ / Laravel MySQL / MariaDB Local disk or S3 Built-in keyword search with tag filters Fixed shelf, book, chapter, and page structure
Wiki.js Node.js PostgreSQL 17 or 18 Local disk, Git, or cloud storage sync Built-in PostgreSQL full-text search Needs PostgreSQL 17+, Node.js 26, and a dedicated hostname
Outline Node.js / React PostgreSQL & Redis Local disk or S3-compatible storage Postgres full-text indexing Needs an SSO, OIDC, or SAML sign-in provider; BSL 1.1 license
DokuWiki PHP None (Flat-file) Local filesystem directories Internal flat-file text index Page locks to prevent edit conflicts
Docmost Node.js / NestJS PostgreSQL & Redis Local disk, S3, or Azure Blob Database full-text search AI, MCP, API, and SSO in paid editions

The Hidden Costs of Self-Hosting: Infrastructure, Maintenance, and Security

Open-source software licenses do not charge subscription fees, leading many technical managers to assume self-hosting is free. In practice, the total cost of ownership (TCO) for a self-hosted knowledge base is dominated by operational maintenance hours, server infrastructure provisioning, and continuous security hardening.

When an organization assigns a systems engineer or DevOps specialist to maintain internal documentation servers, it diverts expensive technical talent away from core product development.

Maintaining a self-hosted documentation server requires continuous administrative intervention across four distinct operational domains:

  1. Server Compute and Persistent Storage: A production-grade deployment cannot run on an unmonitored desktop computer under someone's desk. It requires virtual machines or cloud compute instances with dedicated CPU, RAM, high-performance SSD storage volumes, and snapshot backup storage.
  2. Database Administration and Upgrades: Relational databases require periodic schema migrations, index optimization, connection pooling adjustments, and point-in-time recovery testing to prevent catastrophic data loss during power outages or disk failures.
  3. Network Boundary Hardening and SSL Management: To allow remote or hybrid employees to reach the documentation server safely, administrators must configure reverse proxies, maintain automated SSL certificate renewals, and establish private network tunnels or VPN gateways.
  4. Vulnerability Patching and Dependency Upgrades: The underlying operating system, container runtime, web server, application dependencies, and database engines require frequent security patches to mitigate remote code execution and privilege escalation vulnerabilities.

Monthly Operational Maintenance Responsibilities

Teams running self-hosted documentation infrastructure must commit engineering time to a recurring maintenance schedule:

  • Weekly OS and Container Patching: Applying operating system kernel patches, updating container images, and restarting services during scheduled maintenance windows.
  • Database Backup Verification: Running automated daily database dumps to remote storage targets, followed by monthly test restores to confirm backup archives are not corrupt.
  • SSL Certificate Lifecycle Management: Monitoring Let's Encrypt certificate auto-renewals on reverse proxies and resolving domain DNS verification failures before public certificates expire.
  • Disk Space and Log Monitoring: Purging application log files, vacuuming database tables, and resizing persistent volume mounts as uploaded document attachments accumulate.
  • User Lifecycle Governance: Provisioning accounts for new hires and revoking access for departed staff, by hand or through an LDAP, SAML, or OIDC connection that someone has to maintain.

Why Retrofitting AI Search onto a Self-Hosted Wiki Is Hard

As modern engineering organizations adopt AI tools, technical leads frequently attempt to retrofit AI search capabilities onto existing self-hosted wikis. Most of these platforms index pages for keyword search. Where AI features exist they vary by product: Docmost includes AI answers in its paid editions, while Wiki.js's documentation says the wiki itself runs no model and embeds nothing, leaving that to outside AI tools that connect to it. Teams that want semantic search across both the wiki and their working files usually end up building it.

Attempting to build a custom retrieval-augmented generation (RAG) pipeline on top of a self-hosted wiki introduces immense technical friction:

  • Custom Ingestion Pipelines: The team must write custom scripts to poll the wiki database or API, extract page contents, strip HTML tags, and chunk text into semantic fragments.
  • Vector Database Operations: The team must provision and maintain a vector store, such as Qdrant or Milvus, or the pgvector extension for PostgreSQL, adding another distributed stateful service to their infrastructure stack.
  • Embedding Model Hosting: The organization must run embedding models on its own hardware or pay recurring third-party API fees to generate vector representations for every document revision.
  • Citation and Source Mapping: The pipeline has to carry page IDs and section anchors through chunking so each generated answer links back to its source; without that mapping, users cannot check an answer against the page it came from.

These custom pipelines become brittle internal software projects that demand ongoing engineering maintenance and introduce new security vulnerabilities.

Fastio features

Turn Company Documentation into an Active Knowledge Base

Unify your team files, automated semantic search, and AI agent access in an intelligent workspace without running infrastructure. Start your 14-day free trial on any monthly plan.

How AI Agents Reach a Self-Hosted Wiki

The modern technical workplace is shifting rapidly from human-only interfaces toward collaborative environments where human engineers work alongside autonomous AI agents. Developers use tools like Claude Code, Cursor, Codex, and specialized autonomous agents to research codebases, draft deployment plans, and troubleshoot production incidents.

For an AI agent to operate effectively, it must have real-time access to current company knowledge: architecture decision records, API contracts, deployment runbooks, and configuration guides.

Self-hosted wikis can serve agents. BookStack has a built-in REST API for pages, books, search, and tags; Wiki.js 3 ships an optional MCP server; Outline includes a built-in MCP server in each workspace; and Docmost lists MCP among its paid features. What remains is the work of running that access: the endpoint lives on the team's own server, someone has to issue and scope the tokens agents use, and the agent reaches the wiki's pages rather than the working files kept in drives and repositories.

Where Web Scraping Fails

When a wiki has no API or MCP access turned on, developers often fall back to having the agent scrape the wiki's web interface. This approach fails for several concrete reasons:

  • Session and Authentication Walls: Autonomous agents running in terminal sessions or IDE extensions struggle to maintain stateful web sessions, solve browser challenges, or authenticate through internal VPNs and reverse proxies.
  • Brittle HTML Extraction: Web scrapers parse rendered HTML markup rather than raw data. A minor update to the wiki's CSS styles, theme templates, or page layout breaks the scraper, causing the agent to ingest navigation headers, footers, and scripts instead of technical instructions.
  • No Version Signal: Scraped HTML gives the agent no structured version or last-updated field, so it cannot reliably tell whether a page is current. Wiki APIs and MCP servers return that metadata; Wiki.js's MCP server, for example, includes each page's last-updated date and version with every result.
  • Read-Only by Nature: Scraping cannot save anything. An AI agent that completes an architecture analysis or generates an incident summary has to write its output back through the wiki's API or MCP server, under a token someone has to issue, scope, and rotate.

Model Context Protocol as the Standard for Agent Integration

Autonomous AI agents require a standardized, protocol-driven method to discover, query, and modify company documentation. The Model Context Protocol (MCP) has become a widely adopted open standard for connecting AI applications to data sources and tools.

Under MCP, an intelligent storage system exposes standardized tools for reading files, executing semantic searches, retrieving source citations, managing file locks, and writing persistent assets. Instead of parsing arbitrary HTML, the agent communicates over Streamable HTTP using well-defined JSON-RPC messages.

Connecting an AI agent to documentation via MCP provides distinct technical capabilities:

  • Targeted Hybrid Search: The agent executes queries that combine exact keyword matching for specific function names, error codes, and server hostnames with semantic vector search for conceptual inquiries.
  • Source-Grounded Citations: Search results deliver document identifiers, the page where one applies, and text snippets, allowing the agent to cite checkable evidence for its conclusions.
  • Bidirectional Read and Write: The agent inspects background documentation, processes technical tasks, and saves updated specifications or collaborative notes directly into the workspace.
  • Deterministic Permission Scoping: The agent works within what its credentials allow: the access of the person or account it acts for, narrowed by the scopes granted to the connection.

How an Intelligent Workspace Replaces the Self-Hosted Wiki

Rather than maintaining a standalone self-hosted wiki that drifts from working reality and needs its own agent connection to run and secure, technical teams can use an intelligent workspace as a centralized company brain.

An intelligent workspace integrates file storage, automated hybrid search, collaborative notes, and native agent connectivity into a unified operational layer. Instead of treating documentation as a separate manual chore that developers avoid, documentation lives directly beside the working files your business runs on: specifications, contracts, architecture decisions, onboarding guides, and agent outputs.

By uniting structured documents and unstructured working files in a single indexed environment, teams give both human colleagues and autonomous agents a single source of truth. Teams have no servers to patch, no database migrations to run, and no vector search pipelines to maintain.

Fastio is a hosted cloud service. It does not run on your own servers, on-premises, or in an air-gapped network, so it does not meet a requirement for physical custody of the servers or for network isolation.

How an Intelligent Workspace Solves the Knowledge Silo Problem

An intelligent workspace operating on Fastio addresses the core failures of self-hosted wikis through purpose-built capabilities:

  • Automatic Indexing on Arrival: When documents arrive in a workspace with Intelligence enabled, the platform indexes them automatically for full-text keyword matching, semantic vector retrieval, and AI chat. Teams never need to configure external vector databases, manage embedding models, or write custom chunking pipelines.
  • Native Cloud Synchronization: Instead of forcing teams to manually re-type cloud documents into a wiki, Cloud Sync connects Google Drive, Dropbox, Box, and OneDrive folders directly to a workspace. Synchronization runs one-way or two-way, on a schedule or on demand, with SharePoint document libraries connecting through the standard OneDrive connector. Synced files automatically adopt the destination workspace's role-based access controls.
  • Standardized MCP Connectivity: AI agents connect to the Fastio remote MCP server over Streamable HTTP across dedicated client endpoints: https://mcp.fast.io/mcp/code for coding agents (Claude Code, Cursor, VS Code, Cline), https://mcp.fast.io/mcp/tools for general desktop and web assistants, and, for ChatGPT and Codex, the Fastio plugin as the preferred setup with https://mcp.fast.io/mcp/operations as the custom fallback. Agents execute hybrid searches, read documents, inspect metadata, and write files through standardized tool calls.
  • Collaborative Notes Kept Beside Files: Teams record runbooks, meeting minutes, and architectural decisions in Collaborative Notes stored directly alongside project files, co-editing live with labeled cursors as people and agents type.
  • Structured Data via Metadata Views: With Metadata Views, teams turn unstructured document repositories into queryable databases. AI extracts typed fields across PDFs, spreadsheets, and presentations without manual template configuration.
  • Deterministic Role-Based Governance: Permissions are established per organization, workspace, and share by member role. A workspace has no folder-level or file-level permissions; every workspace member accesses all files in that workspace according to their role. Search and AI chat answer only from the workspaces and shares the asking person or agent belongs to. Because any organization member can join a workspace by default, restrict the join setting on workspaces that hold sensitive material.
  • Comprehensive Activity Tracking: A detailed activity log records changes across the organization, tracking uploads, edits, moves, deletions, membership changes, and share creations for operational oversight.

Architectural Comparison: Self-Hosted Wiki vs. Intelligent Workspace

The comparison table below details the operational differences between managing an open-source self-hosted knowledge base and deploying an intelligent cloud workspace.

Operational Dimension Self-Hosted Open-Source Wiki Managed Intelligent Workspace (Fastio)
Hosting Your own servers, private cloud, or an air-gapped network Hosted cloud service; it does not run on your own servers or on-premises
Initial Deployment Manual server provisioning, container configuration, database setup Instant workspace creation via web interface or API
File Format Support Pages written in the wiki editor, with files attached; search inside attachments varies by product With Intelligence on, indexing for PDFs, spreadsheets, presentations, markdown, and code
Search Capabilities Built-in keyword or full-text search; AI answers in some paid editions (Docmost) or a pipeline the team builds Built-in hybrid search combining exact keyword, semantic vectors, and metadata
AI Agent Integration REST APIs, plus MCP servers in Wiki.js 3, Outline, and Docmost's paid editions, hosted and secured by the team Native remote Model Context Protocol (MCP) server over Streamable HTTP
Cloud Storage Sync Manual copy-pasting or complex third-party file sync scripts Scheduled and on-demand Cloud Sync for Google Drive, Dropbox, Box, and OneDrive
Ongoing Maintenance Regular OS patching, database backups, SSL renewal, security monitoring No servers for the team to maintain; Fastio runs the infrastructure
Access Boundaries Variable per-app user tables; complex LDAP or OIDC integration Organization, workspace, and share role-based access controls

Migrating from a Self-Hosted Wiki to an Intelligent Workspace

Transitioning from an existing self-hosted wiki to an intelligent workspace does not require halting team operations. Technical teams can execute a smooth transition in three structured phases:

  • Phase 1: Export and Ingest Structured Markdown: Export existing wiki pages as markdown files or PDFs. Upload the exported files into a designated Fastio workspace, or put them in a Google Drive, Dropbox, Box, or OneDrive folder and connect that folder with Cloud Sync. With Intelligence on, the workspace indexes the documents as they arrive, making them searchable by meaning and keyword.
  • Phase 2: Consolidate Dispersed Project Files: Connect existing project folders from Google Drive, Dropbox, Box, or OneDrive to centralize specifications, spreadsheets, and slide decks in the same workspace. This eliminates the artificial barrier between written documentation and actual working files.
  • Phase 3: Connect Developer Workflows and AI Agents: Configure developer coding assistants and autonomous agents to connect to the workspace using the Fastio MCP server. Developers querying architecture decisions or deployment procedures from Cursor, Claude Code, or terminal agents can receive checkable, citation-backed answers derived directly from current company files.

Fastio provides transparent pricing based on workspaces, seats, storage, and usage-based AI credits. Monthly plans start with a 14-day free trial that requires a credit card. Plans include Starter, Business, and Enterprise tiers to accommodate growing teams, large file storage, and active agent workflows. Detailed tier specifications, seat allowances, and storage limits are available on the Fastio pricing page.

Sources

References used to verify factual claims in this guide.

  1. A Gartner survey of 4,861 full-time employees, run from September to November 2022, found that 47% of digital workers struggle to find information or data needed to effectively perform their jobs.

Frequently Asked Questions

What is a self-hosted knowledge base?

A self-hosted knowledge base is documentation software deployed on an organization's own on-premises servers or private cloud instances, granting full custody over underlying data. Unlike multi-tenant SaaS documentation platforms, self-hosting requires the organization to manage server compute, database administration, backup routines, security patches, and network routing.

What is the best open-source self-hosted knowledge base?

The best open-source knowledge base depends on team requirements. BookStack is best for non-technical teams seeking a structured, book-like hierarchy. Wiki.js is best for developers needing Git synchronization and multi-format editing. Outline, which is source-available under the Business Source License, provides the cleanest modern user experience for real-time markdown collaboration, while DokuWiki offers a lightweight, database-free flat-file architecture.

Is BookStack free to self-host?

Yes. BookStack is open-source software released under the MIT License, meaning its code is completely free to download, modify, and host on your own infrastructure. However, organizations still incur costs for server hosting, storage disks, backup management, and engineering maintenance time.

Why choose a self-hosted knowledge base over SaaS?

Organizations choose self-hosted knowledge bases primarily to maintain strict data custody, comply with data sovereignty regulations, prevent proprietary intellectual property from residing on third-party multi-tenant servers, operate within air-gapped networks, or eliminate recurring per-user SaaS license fees.

How do you add AI search to a self-hosted wiki?

It depends on the wiki. Docmost lists AI answers among its paid features, and Wiki.js 3 and Outline let outside AI assistants search the wiki through built-in MCP servers. Otherwise it means building a retrieval-augmented generation (RAG) pipeline, with background jobs to extract and chunk wiki text, vector embeddings from local or API-based models, a vector store such as Qdrant or PostgreSQL with pgvector, and search middleware to query vectors and format citations.

What is the total cost of ownership for a self-hosted knowledge base?

While the software license carries no fee, the total cost of ownership includes virtual server compute, persistent SSD storage volumes, snapshot backups, domain registration, and reverse proxy networking. The largest expense is engineering maintenance time, as technical staff must dedicate regular monthly hours to operating system updates, database migrations, security hardening, and troubleshooting.

How does an intelligent workspace differ from a self-hosted wiki?

A self-hosted wiki is a web application the team runs on its own servers, where people write pages by hand, so documentation can drift from the files the work actually happens in. An intelligent workspace such as Fastio is a hosted service that unifies actual working files (PDFs, spreadsheets, presentations, code, and markdown notes) in shared workspaces, automatically indexing them for hybrid keyword and semantic search, syncing with cloud drives, and exposing them to AI agents via the Model Context Protocol.

How do AI agents access documentation in an intelligent workspace?

AI agents connect to the Fastio remote Model Context Protocol (MCP) server over Streamable HTTP at dedicated endpoints like `https://mcp.fast.io/mcp/code` or `https://mcp.fast.io/mcp/tools`. Through standardized tool calls, agents perform hybrid searches, inspect full documents with citations to the page where one applies, extract structured data, and write updated notes and files directly back to the workspace.

Related Resources

Fastio features

Turn Company Documentation into an Active Knowledge Base

Unify your team files, automated semantic search, and AI agent access in an intelligent workspace without running infrastructure. Start your 14-day free trial on any monthly plan.