Appsmith

From PRIMO.ai
Jump to navigation Jump to search

YouTube ... Quora ...Google search ...Google News ...Bing News


Appsmith is an open-source low-code platform for building internal tools such as admin panels, dashboards, customer support consoles, approval apps, and other business software on top of a company's own databases and APIs. Developers drag prebuilt widgets onto a canvas, connect them to queries against PostgreSQL, MongoDB, REST and GraphQL APIs, SaaS tools, or any of more than 25 supported databases, and write JavaScript wherever they need custom logic, data transformation, or bespoke components. The Community Edition is licensed under Apache 2.0 and can be self-hosted for free, while the paid Business and Enterprise editions add workflows, reusable packages, audit logs, single sign-on, user provisioning, and air-gapped deployment. With more than 41,000 stars on GitHub, Appsmith is one of the most widely followed open-source projects in the internal tools category and is often described as the leading open-source alternative to Retool.

Appsmith was founded in 2019 by Abhishek Nayak, Arpit Mohan, and Nikhil Nandagopal with the goal of letting backend engineers build internal web apps quickly without hand-coding every interface. The company, with offices in San Francisco and Bengaluru, raised a $41 million Series B led by Insight Partners in June 2022, bringing its total funding to $51.5 million, with Accel, Canaan Partners, and OSS Capital among its backers. Its customer stories include GSK, which built an app to patch 3,500 Linux servers in a single sprint, and Block, which reports processing new customer applications 50% faster, and its site lists users such as Twilio, Dropbox, Amazon Web Services, and ByteDance. Reviewers consistently describe it as a developer-first tool that is fast and flexible for people who write JavaScript but less suited to non-technical builders.

Appsmith's relationship with AI has moved quickly and not always in a straight line. It launched a free, managed Appsmith AI datasource in early 2024, then a separate Appsmith Agents product in April 2025 that embedded AI assistants in tools such as Salesforce, Zendesk, Slack, Notion, and Gmail through a Chrome extension. By late 2026 the managed datasource had been retired, Appsmith Agents had stopped accepting new sign-ups, and the company had refocused AI on the builder itself with an Ask AI assistant that writes SQL and JavaScript, a copilot that generates custom widgets, and an Appsmith MCP Server, in beta since September 2026, that lets outside AI agents such as Claude Code, Cursor, and Codex build and edit Appsmith apps under a user's own permissions. Apps themselves can still call any large language model through OpenAI, Anthropic, Google AI, and other datasources using the customer's own API keys.

That path makes Appsmith a useful case study in how low-code platforms are adapting to AI agents. Rather than replacing its visual builder with prompt-to-app generation, Appsmith is exposing the builder, its data connections, its permissions, and its Git history to agents through structured tools, with human approval required for anything destructive. The approach assumes that internal tools still need governed data access, review, and deployment discipline even when an agent does much of the building.


Appsmith gives developers a fast, open-source way to put a governed interface on their own data, and in 2026 it began handing that same builder to AI agents, with every change made under a real user's permissions and every destructive step waiting for a human to approve it.


Mechanisms

An Appsmith application is a set of pages built from widgets, such as tables, forms, charts, lists, inputs, and buttons, that a developer drags onto a canvas and configures in a property pane. Widgets are wired to queries and APIs through datasources, and any widget property can be bound to data or logic with JavaScript expressions written inside double curly braces, so that a table, for example, displays a query's results. Reusable logic lives in JS Objects, which hold variables and functions in a central code editor with autocomplete, linting, and debugging, and developers can import external JavaScript libraries or build custom widgets in HTML, CSS, and JavaScript when the built-in widgets are not enough. When a user clicks a button or submits a form, events trigger queries, JavaScript functions, navigation, or alerts, and the results flow back into the bound widgets.

Appsmith connects to data rather than storing it. Its GitHub repository describes integrations with more than 25 databases and any API, and the documented connectors span relational and document databases, REST and GraphQL APIs, SaaS tools, and model providers such as OpenAI, Anthropic, and Google AI. Datasource credentials are stored on the Appsmith server rather than in the app, and datasources can be configured per environment so the same app can point at development, staging, and production data.

Applications are versioned with Git. Developers can connect an app to a repository, work on branches, review changes through pull requests, and, on the Enterprise plan, use continuous deployment so that merged changes on designated environment branches are deployed automatically. Finished apps are published to users under role-based access control, can be embedded in other websites or products, and run either on Appsmith Cloud or on infrastructure the customer controls, including Docker, Kubernetes, and air-gapped environments.

Workflows, available on paid plans, add server-side automation to the same environment. A workflow is a sequence of steps orchestrated in JavaScript that can call queries and APIs, and it can pause at an approval step until an authorized user in the workspace approves or rejects the request from an Appsmith app, with access controls and audit trails around each decision. Appsmith originally ran workflows on the Temporal engine and has been moving them to a native engine, and v2.6.0 in October 2026 told Enterprise users to finish all Temporal runs before upgrading.

The Physics

Appsmith's strengths and limits both follow from its developer-first design and from running against live company data. The constraints that matter most concern who can build, how apps perform, what they cost, and how far AI can be trusted with the builder. The main constraints are:

  • JavaScript is the escape hatch and the barrier: Anything beyond basic configuration is written in JavaScript, which gives developers fine control but leaves non-technical builders facing a steep learning curve, the most common complaint in user reviews.
  • The browser does a lot of work: Apps evaluate bindings and JavaScript on the client, and reviewers report performance lags with large datasets and manual layout work to make apps usable on mobile screens.
  • Security is shared with the data source: Appsmith enforces roles and permissions on apps and pages, but each datasource connection carries the access of its stored credentials, so least-privilege database accounts matter. Appsmith's 2026 releases repeatedly tightened datasource authorization, protection against server-side request forgery, and Git import handling.
  • Per-user pricing: Every member added to a workspace is a billable user on Business, at $15 per user per month for up to 99 users, and Enterprise starts at $2,500 a month for 100 users, with no tier in between.
  • Self-hosting is free but not effortless: The Community Edition costs nothing, but the operator runs MongoDB, Redis, upgrades, backups, and security patches, and features such as workflows and SAML single sign-on add a PostgreSQL dependency.
  • AI acts with a real user's power: An MCP key authenticates an AI client as the user who created it, with all of that user's permissions and no more, so an agent misled by a prompt injection or a bad instruction can do anything that user could do, except that destructive steps wait for explicit confirmation.
  • Vendor-managed AI can disappear: The retirement of the managed Appsmith AI datasource broke existing queries unless teams rebuilt them by hand on bring-your-own-key provider datasources.

These limits explain how successful teams use Appsmith. A few developers build and maintain apps, often with Git and code review, while a larger group of operations, support, and finance staff use them, and sensitive access is controlled in the database as well as in the app. AI features are most useful where a developer reviews what the assistant or agent produced before it reaches users.

Why This Matters

Internal tools are some of the least glamorous and most widely needed software in any company, and Appsmith's history tracks how that work is changing. The company began by giving developers a faster way to build such tools and is now testing how much of that work agents can take on. Its significance shows up in several ways:

  • Open source in a commercial category: Appsmith's Apache 2.0 Community Edition lets organizations inspect the code, self-host on their own infrastructure, and avoid lock-in, which reviewers repeatedly cite as the main reason to choose it over closed platforms such as Retool.
  • A builder exposed to agents: The Appsmith MCP Server lets AI clients such as Claude Code, Cursor, Codex, and Claude build and edit apps through structured tools instead of writing raw widget definitions, SQL, JavaScript, or bindings, with the server compiling each change under the caller's permissions. Its prepare-then-confirm step for destructive actions and its separate Git branches for agent commits make it a concrete example of AI governance for agentic software development.
  • Bring your own model: After retiring its managed AI service, Appsmith points customers to direct OpenAI, Anthropic, and Google AI datasources for apps and to administrator-configured Claude, OpenAI, Azure OpenAI, or locally hosted Ollama models for Ask AI, so organizations decide where their prompts and data go.
  • The economics of free AI: According to the pull request that began its retirement, Appsmith paid for the language model tokens behind the free Appsmith AI service, and the AI graveyard site Killed by AI attributes the shutdown to unsustainable costs.
  • Human-in-the-loop by design: Workflows were built around approval steps that pause automation until a person decides, a pattern Appsmith itself suggested for gradually handing more decisions to AI while keeping a human with the final say.

Appsmith competes with Retool, Budibase, ToolJet, and a growing set of AI app builders that generate whole applications from a prompt, as well as the other platforms on the Prompt-to-Apps page. Its distinctive position combines an open-source core, Git-based development, self-hosting down to air-gapped networks, and an agent interface that works through the same permissions as human developers. Whether that combination is enough against prompt-first builders is one of the open questions for the category.

Building Blocks

Component What It Does Notes
Widgets Prebuilt UI components such as tables, forms, lists, charts, inputs, buttons, and modals, plus a Card widget added in 2026. Drag-and-drop layout; apps built with the newer layout system, formerly called Anvil, are now called responsive apps.
Custom widgets Developer-built components written in HTML, CSS, and JavaScript, with access to external libraries. A built-in AI copilot can generate, edit, and debug custom widget code.
Datasources Connections to more than 25 databases, REST and GraphQL APIs, SaaS tools, and model providers. Credentials stay on the Appsmith server; Business adds premium integrations and Enterprise adds custom integrations.
Queries and JS Objects Queries read and write data, and JS Objects hold reusable variables and functions in a central editor with autocomplete, linting, and debugging. Since v2.2, APIs, queries, and JS Objects can be copied between applications.
Bindings and events JavaScript expressions in double curly braces connect widget properties to data, and events such as clicks or submissions run queries and functions. Most of an app's behavior is defined here.
Reusable packages Shared libraries of queries and JavaScript that several apps can use. Business and Enterprise.
Workflows Server-side automations orchestrated in JavaScript, including human-in-the-loop approval steps that apps can trigger, list, and resolve. Business and Enterprise; moving from Temporal to a native engine in 2026.
Git and CI/CD Connects apps to Git repositories for branching, pull requests, and version history. Free allows 3 repositories, Business unlimited, and CI/CD requires Enterprise.
Access control and identity Roles, groups, and granular permissions, with Google and GitHub sign-in, SAML and OIDC single sign-on, and SCIM provisioning. Custom roles and audit logs on Business; SAML, OIDC, and SCIM on Enterprise.
Embedding Publishes apps publicly or embeds them in other sites and products. Private embedding with shared logins requires Enterprise.
Deployment Appsmith Cloud, or self-hosted on Docker, Kubernetes, or an AWS machine image. Enterprise add-ons include an air-gapped edition and dedicated managed hosting.
Ask AI In-editor assistant that writes, explains, fixes, and optimizes SQL and JavaScript, using the connected schema as context. Administrator-configured provider: Claude, OpenAI, Azure OpenAI, or Ollama; in the Community Edition since v2.3.
Appsmith MCP Server (BETA) Remote endpoint that lets AI clients list, build, edit, deploy, and roll back apps through structured tools. Off by default; per-user keys valid 30 to 365 days; destructive steps need confirmation.
Appsmith Agents (Legacy) AI assistants embedded in tools such as Salesforce, Zendesk, Slack, Notion, and Gmail through a Chrome extension. No longer accepting new sign-ups; existing users keep access.
Appsmith AI (Retired) Free, managed AI datasource for text generation, summarization, classification, entity extraction, and image classification and description. Stopped working on September 30, 2026.
Plans A free cloud plan, Business priced per user, Enterprise from 100 users, and the free open-source Community Edition. Free covers up to 5 users; Business costs $15 per user per month for up to 99 users; Enterprise costs $2,500 per month for 100 users.

AI Offerings

Appsmith has shipped several distinct AI products since late 2023, and they differ in who uses them and whether they still exist. Some help developers build apps, some put AI inside the apps those developers ship, and two have already been wound down. The sections below cover each offering in turn.

Ask AI

Ask AI is an assistant built into the Appsmith editor that writes and refines code from natural-language prompts. In the SQL query editor it can write new SELECT, JOIN, and filter queries, add pagination, fix syntax, and optimize slow queries, using the connected datasource's schema as context, and in JavaScript modules and JS Objects it can generate functions, transform API or query results, and refactor, explain, or fix selected code. An administrator turns it on in the AI Assistant settings and chooses a provider, either Claude from Anthropic, OpenAI, Azure OpenAI, or a locally hosted Ollama model, then supplies the API key and model, with claude-sonnet-4-6 as the documented default. Ask AI requires Appsmith v2.0 or later, reached the Community Edition with v2.3 in August 2026, and is not offered on the free hosted version at app.appsmith.com.

Custom Widget Copilot

The custom widget copilot helps developers create the components that Appsmith's built-in widgets do not cover. Introduced during an Appsmith launch week in early 2025, it adds an AI tab to the custom widget builder where a developer describes a widget in plain language and receives generated HTML, CSS, and JavaScript, which can then be edited by hand in the regular code tabs. It can also modify existing custom widgets and suggest fixes when a developer describes a bug, and v2.3 in August 2026 added native AI copilot capabilities directly to custom widgets.

LLM Datasources in Apps

Apps built on Appsmith reach AI the same way they reach any other service, through datasources. Native datasources for OpenAI, Anthropic, and Google AI let a query send a prompt, along with data from widgets or other queries, to a model using the customer's own API key, and other models with an HTTP API can be called through the REST API datasource. Common uses documented by Appsmith include drafting responses to support tickets, classifying customer reviews by sentiment, recommending decisions such as loan approvals or refunds for a human to confirm, and extracting data from receipts and identity documents. Appsmith's own guide to retrieval-augmented generation shows how to pass reference files or images with a prompt so the model answers from company material.

Appsmith MCP Server (BETA)

The Appsmith MCP Server, introduced as a beta in v2.4 on September 9, 2026, turns Appsmith's builder into a set of tools that outside AI agents can call over the Model Context Protocol (MCP). It runs as a remote endpoint on the customer's own Appsmith domain, and documented clients include Cursor, VS Code, Windsurf, Claude Code, the Codex CLI, Claude.ai and Claude Desktop through custom connectors, and the ChatGPT desktop app inside Codex threads. Agents never write raw widget definitions, SQL, JavaScript, or bindings themselves, because they call tools such as building an application or adding a query and the server compiles each change under the calling user's own permissions. The setup guide covers Appsmith Business Cloud and self-hosted Enterprise deployments, and the server stays off until an administrator enables it.

Its safety model rests on identity and approval. Each user creates personal keys that authenticate the client as that user and grant no extra permissions, valid for 30, 60, 90, 180, or 365 days, with at most 10 active keys per user. Reads and non-destructive edits to an app's unpublished version run without a prompt, while deletions, deployments, rollbacks, Git commits, and data-changing actions use a two-step prepare and confirm flow that the user must approve. For Git-connected apps, agent work goes to branches in an mcp/ namespace with an [mcp] commit prefix, and Appsmith warns operators that those pushes will trigger any CI pipelines or webhooks watching the repository.

Appsmith expanded the server quickly after launch. Version 2.4.3 moved MCP sessions into Redis so they work across replicas, v2.5.0 added structured JavaScript objects, more MongoDB filters, widget setting patches, JavaScript function wiring, and layout checks, and v2.5.1 and v2.6.0 extended structured SQL filters and governance audit tools while preventing MCP-authenticated Git requests from checking out remote branches or tags. Because the agent acts as a real user, Appsmith's documentation recommends least-privilege practices when issuing keys.

Appsmith Agents (Legacy)

Appsmith Agents was a separate product, launched in April 2025 around Google Cloud Next, that brought AI assistants to customer-facing and business teams in sales, support, customer success, and human resources. Delivered through a Chrome extension, the agents worked inside tools such as Salesforce, Zendesk, Slack, Notion, and Gmail, where they could answer questions about company data drawn from several linked sources and launch automated workflows without the user switching tabs. Appsmith described the idea as continuous context, giving models access to business data without fine-tuning or complex RAG pipelines. By late 2026 the Appsmith Agents site was labeled legacy and no longer accepted new sign-ups, with existing users keeping access while Appsmith integrates agent capabilities into an upcoming update.

Appsmith AI (Retired)

Appsmith AI was a managed AI datasource, introduced in early 2024, that let developers add text generation, summarization, classification, entity extraction, and image classification and description to apps without supplying any API keys. Developers could also upload files such as a knowledge base or standard operating procedures at the datasource level so that responses drew on internal documents. Appsmith paid for the model tokens behind the service and moved to shut it down in July 2026, after which v2.3 blocked new Appsmith AI datasources and queries in August and every existing one stopped working on September 30, 2026, whether or not an instance had been upgraded. Queries were not migrated automatically, so teams had to rebuild them on OpenAI, Anthropic, or Google AI datasources with their own keys.

Getting Started

Appsmith can be tried in minutes on its cloud service or installed on a server you control. The fastest path is to connect a real data source, build one small screen, and add Git, permissions, and AI only once the app does something useful. The steps below follow Appsmith's documentation and pricing page.

  1. Choose cloud or self-hosted: Sign up for Appsmith Cloud, which is free for up to five users and starts with a 15-day Business trial, or install the open-source Community Edition with Docker, Kubernetes, or the AWS machine image. Self-hosting is free under Apache 2.0, but you run upgrades, backups, MongoDB, and Redis yourself.
  2. Create a workspace and app: From the dashboard, create a workspace for your team and a new application, or start from a template such as a customer support, inventory management, or IT asset tracker app.
  3. Connect a datasource: Add a connection to a database such as PostgreSQL or MongoDB, a REST or GraphQL API, or a SaaS tool, and test it. Use a database account with only the permissions the app needs.
  4. Write queries: Create queries that read and write data, using parameters bound to widget values instead of hard-coded text.
  5. Build the interface: Drag widgets such as a Table, a Form, and a Button onto the canvas and bind their properties to query results with JavaScript expressions in double curly braces.
  6. Add logic in JS Objects: Put reusable functions in JS Objects, wire them to widget events such as button clicks, and import JavaScript libraries if needed.
  7. Turn on Ask AI (optional): An administrator can enable the AI Assistant, choose Claude, OpenAI, Azure OpenAI, or Ollama, and enter an API key, after which builders open Ask AI from the query or JavaScript editor or by typing /ask AI. Review every suggestion before running it.
  8. Connect Git: Link the app to a repository, work on branches, and review changes through pull requests before merging.
  9. Set permissions and publish: Invite users, assign roles, and deploy the app, keeping in mind that on paid plans every workspace member is a billable user.
  10. Add workflows (Business and Enterprise): Create a workflow for background automation or approvals, then trigger it and resolve its approval requests from an app with workflow queries.
  11. Connect an AI agent (optional): On Business Cloud or self-hosted Enterprise, an administrator enables the MCP Server (BETA), each builder creates a personal key, and the key is added as a bearer token in an MCP client such as Claude Code or Cursor pointed at the instance's /mcp endpoint.
  12. Keep current: Self-hosted operators should watch the release notes, which in 2026 included many security fixes and breaking upgrade notes, and upgrade promptly.

A few habits prevent most problems. Keep secrets in datasources rather than in widget properties or JavaScript, give AI agents keys from accounts with the least access that will do the job, and review agent changes on their mcp/ branches before merging. Teams with no one to maintain servers should weigh the per-user cost of Appsmith Cloud against the operational cost of self-hosting.

Examples

Appsmith's case studies, templates, tutorials, and community projects describe a wide range of apps, and the examples below are drawn from those sources rather than invented. Most are internal tools for operations, support, finance, and IT teams, built by a few developers for a larger group of users. AI features appear both in how the apps are built and in what they do.

Operations and IT

GSK used Appsmith to build an app that patched 3,500 Linux servers in a single sprint, an example of the IT automation tools Appsmith promotes for putting interfaces on top of workflow scripts. Appsmith's templates and use cases include admin panels, back-office apps, CRUD front ends, dashboards, portals, inventory management, and IT asset trackers, and its YouTube channel walks through a CI/CD workflow management dashboard. Teams often start with one of these patterns and connect it to their own databases and APIs.

Support and Customer Operations

Customer support is one of Appsmith's main solution areas, with dashboards that bring customer records, orders, and tickets onto one screen for support agents. Block reports processing new customer applications 50% faster with Appsmith, and HeyJobs says it adds features to its campaign and budgeting apps 90% faster. Appsmith's own tutorial builds an AI-powered support dashboard that retrieves customer reviews and classifies their sentiment, a pattern that now runs on a direct OpenAI, Anthropic, or Google AI datasource.

Approvals and Compliance

Workflows make approval apps a natural fit. Appsmith has published guides for a Stripe refund request manager that holds refunds for approval before they execute and for its own GDPR data deletion process, in which a senior customer support manager approves each deletion request before the workflow removes data across MongoDB, Segment, Mixpanel, and other systems, and it reports customers building procurement workflows that evaluate vendors while decision makers keep the final say. Earlier examples from the workflows beta included order refund and loan sanction approvals that route through several teams.

AI-Powered Apps

Appsmith reported more than 150 generative AI apps built on the platform in the three months before December 2023, and its community has documented many patterns since. Tutorials show apps that generate tags and product descriptions for e-commerce listings and save them back to Airtable, a reusable chat widget that works with any AI chat API and adds Markdown formatting and code highlighting, and PDF reports whose summaries are drafted by a language model from dashboard data. A third-party tutorial adds the Weavy AI copilot to an Appsmith app so users can ask questions about a table of data.

Agent-Built Apps

With the MCP Server, a developer can ask an agent in Claude Code or Cursor to list workspaces, read a page, build a new application, add queries, or roll back a change, and the agent does so through Appsmith's tools rather than by editing files directly. Destructive steps pause for the developer's approval, and Git-connected changes land on an mcp/ branch for review. Community developers have also published unofficial MCP servers that drive the Appsmith editor in a browser or call Appsmith's REST API, one of them noting that no official server existed when it was written.

Sample Requests

The table below shows illustrative requests built from documented capabilities, with the feature each uses and the plan it typically needs. Wording matters less than stating the data, the result, and any limits. Plan details reflect Appsmith's documentation and pricing page as of late 2026.

Request Where It Runs Plan Needed
"Write a query that returns the 20 most recent orders for the selected customer, with pagination." Ask AI in the SQL editor Administrator-configured AI provider; not on the free hosted app.appsmith.com
"Turn this API response into rows for the orders table and format the dates." Ask AI in a JS Object Administrator-configured AI provider
"Build a Kanban board widget with drag-and-drop columns for ticket status." Custom widget copilot Cloud and self-hosted
"Classify each new review as positive, negative, or neutral and store the label." Query on an OpenAI, Anthropic, or Google AI datasource Any plan, with your own API key
"Hold any refund over $500 for a manager's approval before calling Stripe." Workflow with an approval step Business or Enterprise
"Build an admin app for the users table with search, edit, and delete, then commit it to Git." Agent through the Appsmith MCP Server (BETA) Business Cloud or self-hosted Enterprise

Known Limitations

  • Developer skills required: Reviews agree that non-developers struggle, because real apps need JavaScript and an understanding of queries and bindings.
  • Performance and layout: Users report lags on large datasets, extra work to make layouts responsive on mobile, and templates that sometimes ship with errors.
  • Pricing cliff: Business tops out at 99 users and Enterprise starts at $2,500 a month for 100 users, while every workspace member, including builders, counts as a user.
  • Free cloud limits: The free cloud plan allows up to five users, five workspaces, and three Git repositories, with no workflows, packages, audit logs, or custom roles.
  • AI churn: Appsmith AI was retired in 2026 and Appsmith Agents closed to new users, so teams that built on either had to change course, and the MCP Server is still a beta whose tools and configuration may change.
  • Agent permissions: MCP keys carry the full permissions of their owner, the server uses header-based keys rather than OAuth, and regular ChatGPT conversations and connectors are not supported.
  • Ask AI availability: Ask AI needs an administrator to supply a provider and key and is not available on the free hosted version at app.appsmith.com.
  • Self-hosting burden: Self-hosted instances need regular upgrades, and 2026 releases brought breaking changes such as a server schema update for Git-connected apps, re-entered datasource credentials, and a workflow engine migration that required finishing pending approvals first.
  • No built-in app database: Appsmith connects to external databases rather than bundling one for app data, as some competitors do, so teams must provide and manage their own.

History

Recent Developments

  • 2019: Abhishek Nayak, Arpit Mohan, and Nikhil Nandagopal founded Appsmith to help backend engineers build internal web apps quickly with a low-code approach.
  • October 2021: Appsmith closed a Series A round led by Canaan Partners, with Bessemer Venture Partners participating.
  • June 22, 2022: Appsmith raised a $41 million Series B led by Insight Partners, with Accel, Canaan Partners, and OSS Capital participating, bringing total funding to $51.5 million.
  • 2023: Appsmith laid off about 25% of its workforce, roughly 35 employees, according to Inc42.
  • December 2023: Appsmith launched an OpenAI integration after seeing more than 150 generative AI apps built on the platform in three months.
  • February 2024: Appsmith introduced Appsmith AI, a free managed AI datasource that needed no API keys.
  • Early 2025: An Appsmith launch week introduced the custom widget copilot and managed hosting for Enterprise customers.
  • April 2025: Appsmith launched Appsmith Agents around Google Cloud Next, embedding context-aware AI assistants in tools such as Salesforce, Zendesk, Slack, Notion, and Gmail through a Chrome extension.
  • 2025: Appsmith dropped hourly billing in favor of a flat $15 per user per month on its Business plan.
  • July 9, 2026: Version 2.2 added copying of APIs, queries, and JavaScript objects between applications.
  • July 17, 2026: A pull request to retire the managed Appsmith AI datasource noted that Appsmith paid for the language model tokens behind it.
  • August 13, 2026: Version 2.3 blocked new Appsmith AI datasources, brought Ask AI to the Community Edition, added native AI copilot capabilities to custom widgets, and introduced a Card widget.
  • September 9, 2026: Version 2.4 introduced the Appsmith MCP Server in beta.
  • September 30, 2026: Appsmith AI reached end of life, and all remaining Appsmith AI queries stopped working.
  • October 9 and 21, 2026: Versions 2.5.0 and 2.5.1 expanded the MCP tools, updated Git-connected apps to a new server schema, and added a rate limit on self-serve sign-ups.
  • October 28, 2026: Version 2.6.0 began the move to a new workflow engine, requiring Enterprise users to finish runs on the old Temporal engine before upgrading, and patched a Node.js Brotli decoder vulnerability.
  • Late 2026: The Appsmith Agents site, now labeled legacy, stopped accepting new sign-ups while Appsmith integrates agent capabilities into an upcoming update.

Further Resources

Additional Viewing

Load video
YouTube

Building an internal tool in 5 minutes on Appsmith shows a simple admin dashboard being built to look up customers and update their details. It is a quick introduction to the widget, query, and binding model described in The Mechanism.

Load video
YouTube

Announcing Appsmith Custom Widgets introduces custom widgets, which let developers write their own components with their preferred libraries. It provides background for the custom widget copilot described in AI Offerings.

Load video
YouTube

AppsmithAI: Adding Multi-Step AI To Classify Images & Generate Text is a 2024 workshop in which Joseph Petty of Appsmith adds the built-in Appsmith AI to an existing app to generate tags and descriptions for e-commerce products and save the results to Airtable. The Appsmith AI datasource has since been retired, but the same pattern works with OpenAI, Anthropic, or Google AI datasources.

Load video
YouTube

Appsmith BUILD - Demo by Co-Founder Nikhil Nandagopal comes from Appsmith's BUILD event, where the founders and team discussed real-world business needs for custom apps and best practices for low-code development. Co-founder Nikhil Nandagopal demonstrates what was new in the platform at the time.

Load video
YouTube

How to Self Host Appsmith Full Guide 2026 is a step-by-step guide to installing and running Appsmith on your own server. It complements the self-hosting steps in Getting Started.

References

Retrieved from "https://primo.ai/w/index.php?title=Appsmith&oldid=54916"