Softr
Also known as: softr.io, Softr Agents, Softr AI App Builder, Softr Database AI Agents
Softr is a Berlin based AI app builder for business operations, used for client portals, internal tools, CRMs and intranets on its own databases or on Airtable, HubSpot, Salesforce and SQL sources. A prompt produces the database, pages, user roles and logic, which teams then refine visually. Softr Agents act on the result: they read and change records, send Gmail and Slack messages, update HubSpot and call MCP servers, in chat inside an app or on a schedule, with approval steps on write actions.
Softr sells a hosted platform for building business applications without writing code, from client portals and internal tools to CRMs, team intranets, knowledge bases, inventory trackers and dashboards. Apps run on Softr Databases or on more than 20 outside sources, including Airtable, Google Sheets, Notion, HubSpot, Salesforce, monday.com, Supabase, PostgreSQL, Snowflake and BigQuery, and come with sign in, user groups, permissions and hosting included.
Softr Platforms GmbH runs the company from Berlin, and more than 1 million builders and 7,000 organizations build on it, among them Google, Netflix, UPS, Stripe, Celonis, the NBA, Clay and MIT. Celonis runs a knowledge base for more than 1,500 team members on Softr and saves hundreds of hours a week, and Minerva Network cut its administrative workload by 70% while athlete registrations rose 50%.
The AI app builder takes a plain English description and generates the database with its fields and relationships, the pages and navigation, the user groups and the workflow logic, after which builders keep editing visually or through vibe coding blocks that hold custom code. Softr Agents are a separate layer that acts on the running business. An App agent chats with users inside selected Softr apps and works only with the data and actions that signed in user is allowed. A Scheduled agent runs unattended on a recurring schedule.
Each agent gets plain language instructions, a model chosen in the editor with an optional reasoning effort setting, uploaded or linked knowledge that it searches and cites, and tools: record actions in Softr, Google Sheets and HubSpot, Gmail and Slack messages, Softr's own email service, 17 predefined MCP servers such as Airtable, Notion, Linear, Attio, Apollo and Atlassian, and any custom MCP endpoint. Database AI Agents sit on individual table fields and fill them with summaries, tags, extracted invoice dates or web research whenever a record is created or a field changes. Softr Workflows add triggers on database events, form submissions, button clicks, webhooks, schedules and incoming Gmail, with branches, loops, waits, API calls and JavaScript or Python steps.
Pricing is per workspace with AI credits included. Free carries 5 credits a month, Basic is $19 a month billed yearly ($25 monthly), Pro is $99 ($119 monthly) and Business is $329 ($395 monthly), with Enterprise quoted. AI features pause when credits run out until the next reset or an add on purchase, and higher plans can bring their own model provider key. Softr holds a SOC 2 Type II report, hosts on AWS in Germany or the US, and adds SAML or OpenID single sign on, SCIM provisioning and audit logs on Enterprise.
Softr fits operations, IT and agency teams that keep their data in spreadsheets, Airtable or a CRM and want a portal, an internal tool and a few working agents on one bill, without engineers. It fits less well for teams that need exported source code they can host themselves, agents that operate websites through a browser, or an evaluation suite that scores agent behavior before release.
Vendor details
Canonical URL
https://www.softr.io
Category
Coding agent
Subcategory
AI app builder for business operations, with agents that act on app data in chat or on a schedule
Funding status
Private company. Softr Platforms GmbH, registered in Berlin, with managing directors Mariam Hakobyan and Artur Mkrtchyan.
Company status
independent
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
Data sources include Softr Databases, Airtable, Google Sheets, Notion, HubSpot, Salesforce, monday.com, ClickUp, Microsoft Excel, Supabase, Xano, SmartSuite, Coda, PostgreSQL, MySQL, SQL Server, Snowflake, BigQuery, MariaDB and any REST API. Agents act in Softr, Google Sheets, HubSpot, Gmail and Slack, connect to 17 predefined MCP servers including Airtable, Notion, monday.com, Attio, Apollo, Brevo, Calendly, Linear, Supabase, Clay and Atlassian, and accept custom MCP servers. Workflows add Call API, custom code steps and Zapier and Make connections, and the Softr MCP server lets Claude, ChatGPT, Cursor and Codex build apps, databases and workflows.
In practice
An accounting firm builds a client portal from a prompt on Softr Databases. A Scheduled agent runs every Monday morning, finds clients whose tax documents are still missing, drafts a reminder in Gmail for each and posts the list to the partners' Slack channel.
A property manager adds an App agent to its tenant portal. A tenant asks it to log a broken heater, and the agent prepares the maintenance record under that tenant's own permissions, showing a summary and waiting for the tenant to confirm before the record is saved.
A sales operations team puts a Database AI Agent on the company field of its leads table. Each new lead triggers a web search, the agent writes industry, head count and a one line summary into the record, and a workflow then assigns the owner and posts the lead to Slack.
Sources & related URLs
Related / legacy domains
Research sources
Agentic Index coverage score
12.0 / 14 capabilities · 86%
| Integrations & Tool Calling | Full |
|---|---|
|
Agents act in outside systems through authenticated tools. Native tools read, create, update and delete records in Softr, Google Sheets and HubSpot, send Gmail messages and drafts, and post Slack channel and direct messages, each set to run on the user's own connection or a shared one, with each input either left to the agent or fixed. Seventeen predefined MCP servers, among them Airtable, Notion, monday.com, Attio, Apollo, Linear, Supabase, Clay and Atlassian, hand an agent that service's full tool set, and a custom MCP server can be added with OAuth, a header or no authentication. Workflows reach any REST API through Call API. Sourcedocs.softr.io/agents/overviewread 2026-11-18 |
|
| Workflow Orchestration | Full |
|
Workflows chain steps with branches, conditions and filters, loops over lists and record sets, wait steps, database actions, Call API requests and JavaScript or Python code blocks, and Softr's AI drafts the first version of a flow from a plain description. Wait screens keep an app user informed while a flow they started runs. Inside an agent, the model reads its goal and its always and never rules and picks which tools to call and in what order, so one request can look up a record, update HubSpot and post to Slack without a scripted sequence. Sourcesoftr.io/workflowsread 2026-11-18 |
|
| Knowledge Grounding & RAG | Full |
|
Each agent can carry its own knowledge: PDF, TXT and Markdown uploads or web links that Softr indexes, moving each source through Pending, Indexing and Ready before the agent searches it and cites what it used in a reply. A source that fails has to be removed and added again. Database AI Agents ground field outputs in linked tables and in other fields referenced with @ inside the prompt, read files and PDFs attached to a record, and can search the public web per field, with the prompt naming the sites to draw from. Sourcedocs.softr.io/agents/overviewread 2026-11-18 |
|
| Human Oversight & Guardrails | Full |
|
Any native tool can be set to require confirmation, in which case the agent shows a summary of what it is about to do and waits for the user to approve it. Custom MCP tools take Allow, Ask or Deny permissions one by one, and write actions start on Ask. Scheduled agents run with nobody present, so their native tools cannot ask and Ask tier MCP calls are approved automatically on those runs. App agents act as the signed in user, never with admin rights, and cannot reach a table that no block in the app shows. Sourcedocs.softr.io/agents/overviewread 2026-11-18 |
|
| Security, Identity & Governance | Full |
|
Softr holds a SOC 2 Type II report and shares penetration test results through its trust center on request. Workspace roles decide who can build, launch, change settings and see data, and app user groups control which pages, blocks, buttons and records each group reaches, with global data restrictions filtering rows, IP restrictions and domain restricted sign up. Enterprise adds SAML 2.0 or OpenID Connect single sign on for the studio and for app users, and SCIM provisioning. MCP connections are scoped per workspace with separate access levels for apps, databases and workflows. Sourcesoftr.io/securityread 2026-11-18 |
|
| Observability & Auditability | Full |
|
Every agent run lands in a Runs panel that records the model and provider, duration, each tool call, credits charged, input, cached and output tokens, and the finish reason, with a failed run badge and a filter for errors only. The Test agent panel streams replies and tool calls as they happen. Workflow runs keep execution logs, held for 7 days on Free and longer on paid plans. Application and database audit logs show who changed what and when, and an Audit Logs API lets teams search them from outside. Sourcedocs.softr.io/agents/overviewread 2026-11-18 |
|
| Memory & State Persistence | Partial |
|
An agent keeps relevant details about a user across a conversation, so a request later in a chat builds on what came earlier in it. Nothing carries from one conversation to the next as agent memory with its own scope, retention or delete control. What persists between runs is the business data in Softr Databases or a connected source, which agents read and write as ordinary records, along with the run history kept for review in the Runs panel. Sourcedocs.softr.io/agents/overviewread 2026-11-18 |
|
| Deployment & Data Residency | Full |
|
Customer data sits on AWS in one of two regions, Germany or the US. User data stays in Germany in normal operation, and the US is offered as the alternative region. Apps go live on a Softr subdomain or the customer's own domain and can install on phones as a progressive web app. Apps built on outside sources leave the records where they already live, in Airtable, HubSpot, Salesforce, PostgreSQL, Snowflake or BigQuery, with queries run on Softr's servers so credentials never reach the browser. Sourcesoftr.io/securityread 2026-11-18 |
|
| Prebuilt Agents / Templates / Packs | Full |
|
Database AI Agents come ready made for product descriptions, meeting summaries, enrichment, email follow up, invoice verification, customer feedback, contract review and translation, with a prompt template library behind them. Workflow templates cover new record alerts, weekly AI digests, threshold alerts, web page scraping, onboarding sequences and owner assignment. App templates for ERP, client portals, inventory management and company intranets start a build from a working app with its database and pages rather than a blank screen. Sourcesoftr.io/database-ai-agentsread 2026-11-18 |
|
| Triggers & Channel Coverage | Full |
|
Work starts without a person in several ways. Scheduled agents run unattended on a recurring schedule in a chosen timezone, with an optional opening message for each run. Workflows fire on records created, updated or deleted in Softr Databases, on form submissions, button clicks and record edits inside an app, on inbound webhooks from other systems, on schedules and on incoming Gmail. Database AI Agents rerun when a record is created or a watched field changes, with conditions such as a status value. App agents answer users in chat inside the Softr apps they are attached to. Sourcedocs.softr.io/agents/overviewread 2026-11-18 |
|
| Model Flexibility & Routing | Full |
|
The builder picks the model for each agent in its editor, from several providers grouped into top, middle and lightweight cost tiers, and sets reasoning effort to Off, Low, Medium or High where the model supports it. Database AI Agents offer more than 13 models from OpenAI, Anthropic and Google Gemini, chosen per field, with the cost of each run shown. Higher plans can bring their own provider API key, so usage bills to that provider instead of Softr credits. Sourcedocs.softr.io/agents/overviewread 2026-11-18 |
|
| APIs / SDKs / MCP Extensibility | Full |
|
A REST API takes a personal access token and covers app users (create, invite, deactivate, sync and magic links), the Softr Database API for databases, tables and records, and an Audit Logs API. The Softr MCP server lets Claude, Claude Code, ChatGPT, Codex and Cursor connect over OAuth or a token and create apps, pages, user groups, data restrictions, tables, fields, records, vibe coding blocks and workflows, then preview and launch them, under access levels set per workspace. A published app can also run its own App MCP server for end users' assistants, bound by the app's permissions. Sourcedocs.softr.io/softr-api/api-setup-and-endpointsread 2026-11-18 |
|
| Testing, Debugging & Optimization | Partial |
|
Builders try things before going live. The Test agent panel runs an agent against sample requests and shows each reply and tool call, Database AI Agents preview their output on a sample record before a bulk run, workflows can be test run from the Runs view, and an app can be previewed as a chosen user before it goes out. There are no scored test cases, saved evaluation sets or quality gates that judge an agent's answers against expected results. Sourcedocs.softr.io/agents/overviewread 2026-11-18 |
|
| Browser / Computer-use | Not documented |
|
Agents work through APIs, MCP tools and connected data sources rather than by driving a browser or a desktop. Web research in Database AI Agents runs as a search whose results are written into fields, and a workflow template for web page scraping fetches content without operating the page. Apify sits in the predefined MCP catalog for teams that want its scrapers, but Softr runs no browser of its own for an agent to control, and no agent clicks or types inside another application. Sourcedocs.softr.io/agents/overviewread 2026-11-18 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Pricing
Free $0; Basic $25 a month, or $19 billed yearly; Pro $119, or $99 yearly; Business $395, or $329 yearly; Enterprise quoted
A workspace plan sized by builders, app users, database records, workflow actions and monthly AI credits, with extra builders and app users billed on top
Included quota
Free carries 5 AI credits, up to 3 builders, 5 app users, 5,000 records and 500 workflow actions a month. Basic has 10 credits, 50,000 records and 2,500 actions. Pro has 50 credits, 500,000 records and 10,000 actions. Business has 100 credits, 1 million records and 25,000 actions.
What is public
Softr sells workspace plans with AI credits included. Free costs nothing and carries 5 credits a month. Basic is $25 a month, or $19 billed yearly, Pro is $119 or $99 yearly, and Business is $395 or $329 yearly. Enterprise is quoted after a call with sales.
Billing mechanics
Plans bill monthly or yearly, with two months off for yearly billing. Extra builders, app users, credit add ons and custom domains are added to the plan. Higher plans can connect their own model provider key, so agent usage bills to that provider rather than to Softr credits.
Cost watchouts
Each extra builder costs $5 a month on Basic, $10 on Pro and $20 on Business, and app users past the plan allowance bill monthly from $1 to $2 each. Agent credit use rises with the model chosen, tool calls, knowledge search and output length. Each extra custom domain is $15 a month, or $13 billed yearly.
Variable cost rationale
Each plan has a fixed price, but agent runs draw on monthly credits that scale with model choice and tool use, and app users and builders add charges as a portal grows.
Additional watchouts
Single sign on, SCIM and audit logs sit on Enterprise only.
Overage / add-ons
AI credits reset each month and AI features pause when they run out, until the reset or an add on purchase. Workflows pause at the monthly action limit until it renews or the workspace upgrades.
Sales call required
Mixed (some tiers require a call)
Free / trial
A Free plan with 5 AI credits a month, plus a free trial of paid features with no card required
Lowest paid plan
Basic, $25 a month or $19 billed yearly, with 10 AI credits and one builder
Commercial notes
Pro lets builders work from Claude, ChatGPT or another MCP client on their own AI subscription without spending Softr credits. Enterprise adds SAML or OpenID single sign on, SCIM, audit logs, custom agreements and SLAs, and a customer success manager. Nonprofits and education get 50% off.
Missing data
Softr gives no price for credit add ons and no Enterprise price.
Related vendors
- Cognition — Maker of Devin, an autonomous AI software engineer
- 10Web — Agentic website platform whose AI agents build, host and manage…
- AgentUI — Managed AI app builder for operations teams that generates and hosts…
- Aider — Open source, model agnostic terminal coding agent that edits your…
- Amp — Amp is a coding agent from Amp Frontier Corporation that runs each…
- Anthropic Claude Code — Claude Code is Anthropic's agentic coding tool
Alternatives to Softr
Other products that do the work Softr does, and what each one is.
- Augment CodeAI coding platform built around its Context Engine
- GitLab Duo Agent Platform
- Tabnine
- AmpCoding agent
- CodeRabbitAI code review agent for GitHub, GitLab, Azure DevOps and Bitbucket
- DeepSourceAI code review platform