New features, improvements, and updates to Wire.
Your agent forgets everything between sessions. wire-memory fixes that. It's a free, open source Claude Code plugin that automatically captures decisions, corrections, and coding patterns, then retrieves them when relevant. Memory is stored in a Wire container, so it's portable to any MCP-compatible tool. Install with two commands, connect with one slash command, and your agent stops repeating mistakes. Learn more.
wire_search now returns more relevant results with better ranking. Your agents get higher quality context from every search call, and each call returns fewer tokens, keeping costs low even when the agent explores across multiple turns. See the benchmarks.
Every Wire container now ships with the same 5 tools: explore, search, write, delete, and analyze. No matter how much data is inside. Agents discover your schema and relationships through wire_explore, then query across fields, text, and semantics with wire_search. Write entries, delete them, or trigger analysis to regenerate query tools as your data evolves. Previously a container could generate 45+ tools. Now it's always 5, cutting token overhead by 90% and eliminating the need to reconnect when your schema changes.
Container analysis is now ~40% faster on subsequent runs. As Wire builds a deeper understanding of your context, it gets faster at connecting new information to what it already knows. Relationships between entries are discovered more efficiently, so your containers are ready to query sooner.
Start using Wire without signing up. Add a container to your MCP client and begin uploading files, writing entries, and querying context immediately. Your container runs for 7 days as an ephemeral session. When you're ready to keep it, call the wire_claim tool to generate a claim link. Sign up or log in, and your container moves into your existing organization or a new one — all data, tools, and MCP connections preserved.
The Activity tab in container details now shows richer information about every action. See who committed changes, what type of action was performed, and detailed context for each event. Agent writes, file uploads, analysis runs, and configuration changes are all tracked with full attribution.
When data is returned from tool calls, results now include source attribution showing where each piece of data originated. See which file, agent session, or entry contributed to each result, making it easy to trace information back to its source and verify accuracy.
Container analysis now runs multiple passes over your data, continually recontextualizing entries based on the overall dataset. Each pass builds on the previous one, so relationships between entries are discovered and context is enriched across the entire container. The result is richer, more connected context for your agents to work with.
Rebuilt the file ingestion pipeline for dramatically faster processing. Files that previously took minutes to process now complete in seconds. Upload a CSV, PDF, or any supported file and your content is ready to query almost immediately.
Containers now ship with tools that let agents read and write context directly. Write entries, delete them, batch up to 50 operations at once, or trigger analysis to generate query tools from your data. Use them to capture meeting notes, log experiment results, track customer feedback, or build a knowledge base over time without any file uploads.
A new Activity tab on each container shows a timeline of everything that's happened — file uploads, container updates, visibility changes, pauses, and more. Filter between container lifecycle events and file events, and a blue dot on the tab lets you know when there's new activity you haven't seen yet.
Container initialization, file additions, and file removals are now significantly faster. Flows that previously took over 3 minutes now complete in under 100 seconds. We also improved tool name stability — when you add or remove files, your existing MCP tool names are preserved so connected agents don't break.
Documentation pages now have a "Copy to AI" button. Copy any page as clean markdown or open it directly in Claude or ChatGPT with full context. Great for asking specific questions about Wire without leaving the docs.
Completely redesigned onboarding flow for new users. Create your first container, upload files, and set up your organization all in one guided experience. The new flow includes organization creation with custom slugs and optional billing setup. Anonymous users can now try Wire before creating an account.
Transfer containers between your organizations. All files, tools, and context move with the container. Access this from container settings if you're a member of both organizations.
Containers now get automatic names and descriptions based on their content. When you create a container, it starts with a fun placeholder name. After you upload files and initialize, Wire analyzes the content and generates a descriptive name and purpose-focused description. You can still set your own name and description in container settings if you prefer.
Wire now has a dedicated documentation site here. Find guides for connecting your containers to Claude Code, Claude Desktop, Cursor, Cline, and other MCP clients. Includes quick start guides, authentication docs, and reference material for all available tools.
Improved the resilience and speed of container initialization and update flows. Some operations are now up to 40% faster with better error recovery and retry handling.
Stay ahead of credit depletion with smart notifications. A warning banner appears when your balance drops below your configured threshold, and turns critical when credits run out. Dismiss warnings for 24 hours with smart snooze. If your balance drops significantly, the warning comes back early. Configure your low credit threshold in billing settings to match your usage patterns.
See exactly how your containers are being used with the new Usage tab. Track credits consumed and tool calls at a glance, then browse the full event history with filtering by date range and event type. Color-coded badges distinguish between tool calls and file processing. Timestamps display in your local timezone.
Deleted containers now go to trash instead of being permanently removed. You have 14 days to recover a deleted container with all its files, tools, and context intact. Access your trash from the Containers page to view deleted containers, recover them, or permanently delete them before the retention period ends.
Connect to your Wire containers using OAuth 2.1. MCP clients like Claude Desktop can now authenticate through your browser with a simple authorization flow. Tokens are scoped to specific containers using RFC 8707 resource indicators, so each connection only has access to the container you authorize. Your MCP client needs to support OAuth with resource indicators for this to work. No more copying API keys around.
Control which tools are advertised to AI agents. The redesigned Tools tab lets you list or unlist individual tools with a single click. Unlisted tools are hidden from MCP clients while remaining preserved in your container, and can still be called directly if needed. Connected agents receive instant notifications when tools change via MCP list_changed notifications. Your MCP client or agent needs to support this notification type to see updates without reconnecting.
You can now add new files to containers that are already initialized. Upload additional documents, data, or content and Wire will intelligently merge them with your existing context. New entities are created, existing entities are updated with new records, and your MCP tools are automatically regenerated. Combined with file deletion, you now have full control over your container contents without rebuilding from scratch.
File uploads now continue in the background while you navigate around the app. Start uploading files to a container, then browse other containers or check your settings. Your uploads keep going. A new upload indicator in the header shows progress across all active uploads, grouped by container. Click any upload to jump straight to that container.
Delete multiple files at once with the new batch delete feature. Enter edit mode in the Sources tab, mark files for deletion, and update your container in a single operation. We've also redesigned the container details page with a cleaner tabbed interface—quickly switch between connection info, MCP tools, and source files. Tool counts now appear directly on the tabs so you can see what's available at a glance.
You can now delete individual files from your containers. When you remove a file, Wire cleans up all associated data while preserving everything else to continue operation. No more rebuilding containers from scratch just to remove a file. Your MCP tools are automatically regenerated to reflect the updated contents.
Discover public MCP servers built with Wire. The Explore page showcases ready-to-use context servers you can connect to Claude or other MCP clients instantly. Each server page includes connection instructions, available tools, and example queries. Now featuring UN World Population data, A Christmas Carol, and 146 vintage chocolate recipes from 1914.
MCP tools now share their full parameter schemas with AI agents. When agents list available tools, they receive detailed information about expected parameters including types, constraints, and descriptions. This makes tool invocation more reliable and reduces errors from incorrect parameter usage.
Added support for TXT and TSV (tab-separated values) files. For TXT files, Wire automatically detects whether the content is structured data or plain text, and generates the appropriate tools for your AI agents.
When you delete a container, you now see a progress indicator while deletion is in progress.
Improved how we handle structured data, especially for larger datasets. Container initialization is now faster and more resilient to failures.
Make your containers publicly accessible. Perfect for sharing documentation, research, or datasets with AI agents worldwide. Anyone with the MCP URL can access public containers without authentication. Toggle visibility in container settings to enable or disable public access at any time.
A dedicated settings page for each container. Rename containers, update descriptions, toggle public access, pause or unpause containers, and manage deletion all in one place.
You can now temporarily pause containers to block MCP access without deleting them. When a container is paused, connected AI agents will receive a clear message that the container is temporarily unavailable. Unpause at any time to restore full access. Perfect for maintenance windows or when you need to temporarily disable context sharing.
Added a login button to the navigation bar on the marketing site, making it easier for existing users to access their accounts.
Improved handling for larger files with better progress tracking and error recovery. Files over 1MB are now processed more efficiently.
Introducing Wire: composable context containers for AI agents. Create containers, upload your documents, and connect your favorite AI tools through MCP. Your context, accessible everywhere.