Changelog / shipped work
Product changes,
without the fluff.
Latest updates, improvements, and fixes.
v1.3.1
Agent Memory projects, canonical context, and tier limits
Separate project namespaces, keep stable Rules and Project Summary documents under review, and scale Agent Memory request rates by subscription plan.
AddedSeparate project namespaces inside one Agent Memory source, with optional project scope on each MCP link.
AddedCanonical Rules and Project Summary documents for stable context outside the append-only work log.
AddedRead & Append links can propose canonical replacements for owner/admin review; Full Access can apply them directly for an owner/admin API key user.
ImprovedAgent Memory rate limits now scale across Free, Pro, and Enterprise plans for appends, reads, handoffs, compaction, projects, sessions, and canonical updates.
ImprovedFree supports 20 appends per minute and 60 per hour, Pro supports 120 per minute and 2,000 per hour, and Enterprise supports 600 per minute and 20,000 per hour per user and memory connection.
v1.3.0
Agent Memory: shared project context through MCP
Create hosted Agent Memory connections so compatible AI chats can append structured project updates, search earlier work, and prepare compact handoff context through one remote MCP link.
AddedAgent Memory as a first-class hosted MCP connection type for shared project context.
AddedSix memory tools: register_session, append_memory, search_memory, get_memory_entry, create_handoff_context, and summarize_memory.
AddedStructured append-only entries with Markdown bodies, summaries, types, importance, tags, files, metadata, sessions, and occurrence dates.
AddedPostgreSQL full-text search with filters for dates, tags, files, entry type, importance, and session.
AddedDurable memory summaries with optional source-entry archiving, archived retrieval, and summary-aware handoff context.
ImprovedIdempotency keys make retried memory appends and compaction requests safe from duplicate records.
AddedRead Only and Read & Append MCP-link modes for separating recall from memory writes.
ImprovedMemory inputs redact common secret formats, flag prompt-injection hints, and reject oversized opaque payloads.
v1.2.0
MySQL Connections: hosted MCP access for MySQL databases
Connect MySQL as a first-class database source, discover its schema, and publish authenticated MCP links for supported AI clients without copying the database password into every client configuration.
AddedMySQL connections through a mysql:// URL or structured host, port, database, username, password, and TLS fields.
AddedMySQL schema discovery for tables, views, columns, primary keys, foreign keys, and indexes.
AddedHosted MySQL MCP links with schema tools and single-statement SQL execution governed by link policy and effective MySQL grants.
AddedMySQL query activity and permission-denial records in the existing database activity workflow.
ImprovedDocumentation, setup guidance, source navigation, and database MCP comparisons now cover both PostgreSQL and MySQL.
v1.1.0
OpenAPI Connections: turn any REST API into MCP tools
datamcp is no longer just for databases. Point us at an OpenAPI 3.x specification and AI clients can discover and call its endpoints through a focused MCP toolset, with supported auth injected server-side and access scoped by HTTP method.
AddedOpenAPI connections: paste an OpenAPI 3.x spec URL and datamcp validates it, counts the endpoints, and exposes them to your AI tools over MCP.
AddedWorks with documentation pages too: if your URL is a Redoc or Swagger UI page rather than raw JSON, we extract the embedded spec automatically. No separate JSON endpoint required.
Addedcall_endpoint tool: your AI can make live calls to the API (path params, query strings, and JSON bodies) and get the real response back.
AddedAutomatic auth injection: set an API key or Bearer token once on the connection and datamcp attaches it to every call. We read the API's own securitySchemes and pre-fill the header name for you.
AddedAccess presets for API links: Full Access (any endpoint) or Read Only (GET/HEAD only), so an AI client can read from a production API without being able to create, modify, or delete.
AddedIntrospection tools: list_api_endpoints, get_endpoint_details, and get_api_schema let the AI discover what the API offers before calling it.
v1.0.0
In-App Chat
Chat with your database directly in datamcp. No AI tool, no IDE, no MCP client needed. Ask in plain English, get answers, and export results from the dashboard.
AddedIn-app chat: query any connected database in plain English, right from the dashboard. Answers come as tables, charts, KPIs, or prose depending on what makes sense for your question.
AddedThird-party MCP connections in chat: attach Stripe, Linear, Notion, or any MCP server to a conversation for cross-product queries without leaving the dashboard.
AddedFile attachments: drop an image or PDF via drag-and-drop, paste from clipboard, or use the paperclip button. The AI sees your screenshot or reads your document as context.
AddedData export: ask the AI to export any result set to CSV or PDF. PDFs auto-switch to landscape for wide tables with dynamic column widths.
v0.9.3
Live Activity Indicator
See at a glance how many AI tools and teammates are actually using each connection in real time.
AddedActive clients badge on every connection card - shows how many unique AI tools or users ran a query in the last 15 minutes
AddedHover the badge for a breakdown: what counts as a "client", the rolling window, and the timestamp of the most recent query
ImprovedActivity data lives in Redis, so the counter survives backend restarts and stays consistent across replicas
v0.9.2
Onboarding Polish & Email Preferences
A smoother first-run experience and fine-grained control over the emails you get from us.
AddedEmail preferences in Settings - toggle product update emails on or off anytime
AddedOpt-in checkbox on the last onboarding step, so new users explicitly choose whether they want to hear about new features
FixedOnboarding flow sometimes skipped the final welcome step and dropped users straight into the dashboard
FixedRefreshing the page during onboarding occasionally brought users back to step 1 even after they had completed the survey
v0.9.1
Connection Error Handling
When your database goes down, datamcp now tells you exactly what's wrong and how to fix it - with one-click recovery.
AddedReconnect button for connections in error state - one click re-tests credentials and brings MCP links back online
AddedReal PostgreSQL error messages shown directly on the connection card, so you know exactly what's wrong
AddedEmail alerts to org owners and admins when a connection goes down, with troubleshooting steps
AddedTroubleshooting guide in docs covering the 6 most common connection errors
ImprovedCloud database connections (AWS RDS, Heroku, Supabase) are no longer incorrectly flagged as unreachable due to SSL handshake issues
ImprovedConnection action menu now adapts to state - Reconnect when in error, Pause/Resume otherwise
v0.9.0
AI Schema Descriptions
Your AI tools now understand what your tables and columns actually mean - not just their names and types.
AddedAI-generated descriptions for every table and column, served to your AI tools through get_schema
AddedReview and edit generated descriptions in the dashboard before they reach your AI
Addedresync_schema MCP tool - your AI can re-extract the schema when it detects changes
AddedSchema change history - see diffs between schema versions over time
AddedNew connection wizard with individual fields (Host, Port, Database, User, Password) and SSL mode selector
ImprovedQueries are now up to 10× faster thanks to persistent connection pooling
ImprovedVerified TLS certificates for AWS RDS, Supabase, Google Cloud SQL, Azure, Neon, and Aiven out of the box