AgentSwarms Data and Agent Platform

Run agents on your own infrastructure and ground their work in governed data.

Stars
★ 302
Last updated
today
License
NOASSERTION
Primary language
TypeScript

At a glance

How it runs
Web appSelf-hosted service
Works with
Universal · cross-platformOpenAI API
Cost
Free software; you pay for model usage
Setup effort
High · needs real infrastructure
You'll need
Node.js ≥20.19SupabaseDockerShell / CLINetwork accessLocal filesystemMCP Server
Typical use
A data team moves database and SaaS data into a lakehouse and builds traceable ETL pipelines and downstream SQL models.
Not a fit if
  • Teams requiring automatic multi-region failover
  • Teams seeking to offer the software as a hosted service to others
Website
agentswarms.fyi ↗ set on GitHub
Maintenance
Commits in the last 30 dayslast commit 2026-11-20latest release v1.7.0 (2026-11-14)
Source review
79/100 · Good

What does this agent do, and when should you use it?

AgentSwarms combines Agent Chat, a multi-agent Swarm canvas and RAG knowledge bases with ETL, a DuckDB and Parquet lakehouse, BI dashboards and no-code machine learning. Agents and analysts can query connected sources or the lakehouse; the analyst exposes its SQL, while dashboards and agents can share semantic definitions and access policies. Workflows orchestrate pipelines, model builds, SQL, notebooks, Swarms, dashboard refreshes, approvals and other steps. The platform uses Supabase as its backend and can be deployed with Docker or Kubernetes, with support for multiple model providers. It is aimed at teams that want to self-host data and AI workflows and can provide the runtime, a Supabase project and model credentials.

Users can connect databases, warehouses, files and SaaS sources, build ETL pipelines in a visual canvas or Python mode, and write data to a Parquet lakehouse in their own object storage. The data catalog records table and column metadata and lineage; governed SQL, semantic metrics and row and column policies can be shared across BI, AI Analyst and agents. Users can create RAG knowledge bases, build Swarm graphs, configure tools and run them from the UI, API or a schedule; deployed Swarms checkpoint as they run, and human approval steps can pause execution. AI Analyst breaks questions into steps, writes and runs SQL, checks its work and returns findings with step citations. The platform can also train and register machine-learning models on lakehouse tables, write predictions back, and manage operations through workflows, monitors, audit and traces.

  1. A data team moves database and SaaS data into a lakehouse and builds traceable ETL pipelines and downstream SQL models.
  2. An analyst builds dashboards over row- and column-governed data and uses AI Analyst to query and verify results.
  3. An operations team orchestrates data checks, model retraining, dashboard refreshes and human approvals in one workflow.
  4. An enterprise indexes internal documents and websites in a RAG knowledge base for cited employee Q&A.
  5. A data science team trains classification, forecasting or anomaly-detection models on lakehouse tables and schedules batch predictions back into the lakehouse.

Key setup facts

Pulled from the install and usage notes below. The project README has the current values.

Config files
.env
Runtime
Node.js 20.19
Install notes cover
Docker

How do you install or deploy this agent?

Node.js 20.19 or later is required. For the hosted Supabase path, create a Supabase project and obtain its keys, then install dependencies, configure the environment and apply database migrations. The documentation also provides Docker deployment and a setup script that includes self-hosted Supabase. The full Docker deployment uses about 8 GB of RAM; 16 GB is recommended for model training.

How do you use this agent?

The manual development startup path is shown below. Configure Supabase and model-provider credentials in .env as needed, apply the migrations, then open the local web interface. Docker Compose and Kubernetes deployment are also documented.

What are this agent's strengths and limitations?

Pros
  • Agents, ETL, the lakehouse, BI and machine learning share IAM, audit and governance policies.
  • Lakehouse data is stored as zstd Parquet in user-owned object storage, with snapshot time travel.
  • Supports multiple model providers and exposes an OpenAI-compatible /api/v1/chat/completions endpoint.
  • Swarms, data pipelines, model jobs, approvals and notifications can be orchestrated in one workflow graph.
Limitations
  • Deployment involves Supabase, environment configuration and several services; the full Docker setup uses about 8 GB of RAM.
  • Model features require provider credentials, and call costs depend on the selected provider.
  • The README states that multi-region failover is unavailable; recovery from regional loss requires restoring from backup.
  • The Elastic License 2.0 does not permit offering the software to third parties as a hosted or managed service without a commercial license.

How does this agent compare with similar options?

The README distinguishes this self-hosted repository from the hosted agentswarms.fyi service: the repository is for teams that deploy and run the platform, while the hosted service focuses on courses, labs and a managed learning environment.

Key facts side by side with the most closely related agents.

Agent Source review Form / cost Stars Updated Language Full support on
AgentSwarms Data and Agent Platform This agent 79 · Good Web appFree + model costs ★ 302 today TypeScript OpenAI API
WrenAI GenBI 54 · Major gaps CLIFreemium ★ 18k 2d ago Python Codex · Claude Code
AI Engineering Bootcamp 38 · Major gaps FrameworkFree ★ 959 1mo ago Jupyter Notebook —
Seal Report 54 · Major gaps Desktop appFree ★ 1.6k today C# OpenAI API · Claude API

How does FollowAgents rate this agent?

FollowAgents source review · FARS-2.1
Good
79/ 100 5-point scale 4.0 / 5
This review covers the source as of Nov 13, 2026. The project has shipped 1 stable release(s) since, which have not been re-reviewed. The score and verdict describe the reviewed version only. Recent releases ↓
Trust 21/29
Reliability 11/14
Adaptability 16/18
Convention 15/18
Effectiveness 10/13
Verifiability 6/8
Why each dimension lost points
Trust21 / 29 · 3.6/5

The security model, sandbox, and lakehouse access notes describe concrete permission boundaries, credential flows, SSRF defenses, and audit controls; they also explain user-key handling, encryption scope, and rotation risks. Deductions: the app process holds a service-role key that bypasses RLS, so safety depends on application scoping; the document-generation service skips authentication when DOCGEN_TOKEN is unset (it is intended for private networks); and the sample does not establish full dependency risk coverage beyond a scan gate for fixable critical findings. Reporting and maintenance contact details exist, but publisher identity is unverified.

Reliability11 / 14 · 3.9/5

The docs, CI configuration, and implementation samples are broadly consistent on sandbox refusal, failure reporting, and dependency installation. CI covers type checking, documentation checks, tests, builds, and browser checks. Failure messages score well because the gateway logs rejection reasons and informs clients, and the sandbox explains when it is unavailable. Deductions: the packet cannot cover this large repository, and static evidence cannot establish runtime reliability or consistency across every documented claim.

Adaptability16 / 18 · 4.4/5

The README and installation guide cover local and self-hosted use, multiple operating systems, containers, Kubernetes, and varied workloads. Sandbox documentation states capability boundaries and a known shared-egress-proxy limitation. Deduction: triggers such as schedules and webhooks are described, but the sample does not show full configuration constraints and trigger rules for every capability.

Convention15 / 18 · 4.2/5

The README provides clear feature entry points; the installation guide gives prerequisites, staged setup, and multiple deployment paths; security docs state limitations, operator responsibilities, and a reporting channel. Deductions: the product is broad and the sample cannot establish naming stability across the repository; README and package metadata are supported by a LICENSE file, but ELv2 has a specific hosted-service restriction and is not a permissive general-purpose license; version and maintenance information is present but incomplete, and publisher identity is unverified.

Effectiveness10 / 13 · 3.8/5

The product describes a broad set of Agent and data capabilities, with emphasis on inspectable SQL, usable outputs, orchestration, and auditable governance. Deduction: the scope and operational dependencies are substantial, including Supabase, containers, and optional model services; the sample provides no cost, resource, or outcome data to establish cost-benefit for an ordinary deployment.

Verifiability6 / 8 · 3.8/5

The security docs tie many claims to implementation files, CI specifies automated checks, and installation and sandbox docs give checkable steps and boundaries. Deductions: some security explanations and README feature claims appear only in one source in this packet; this review can distinguish documented claims from visible configuration, but cannot statically confirm that the described behavior is implemented.

Risks and how to mitigate them
  • This low-confidence assessment uses only the supplied file sample; no code was run and product claims were not independently verified.
  • Before deployment, set DOCGEN_TOKEN and restrict network access to the doc-gen service; review service-role paths, customer-key configuration, and the stated scope of data encryption.
  • Elastic License 2.0 restricts providing the software as a hosted or managed service; confirm the license fits the intended use before deployment or redistribution.
Evidence confidence: Low Reviewed Nov 13, 2026 Reviewed revision 41c82d3df37a New commits since this review; the score may not cover them
See the full review method →
Maintain this project? Add the FARS badge to your README

The badge shows the current review score and verdict and links back to this page. It refreshes about once an hour after a re-review.

FARS

[![FARS review](https://followagents.com/badge/agentswarms.svg)](https://followagents.com/en/agents/agentswarms)
<a href="https://followagents.com/en/agents/agentswarms"><img src="https://followagents.com/badge/agentswarms.svg" alt="FARS review"></a>

Recent releases

Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.

  1. v1.7.0 AgentSwarms 1.7.0
    Release notes excerpt
    AgentSwarms 1.7.0 is the first release published as images, and the first that deploys to AWS without assembling it by hand. Every image is built for amd64 and arm64, smoke-tested, scanned and signed. An install picks a release with AGENTSWARMS_VERSION, and rolls back the same way. The schema now installs on a plain PostgreSQL such as Amazon RDS. deploy/aws deploys the whole platform with Terraform: as VMs behind a load balancer, on Amazon ECS, or on EKS from the new Helm chart. Users' code runs on Fargate there, or on a node pool of its own. Behind a load balancer, restarts and rolling …
    View on GitHub ↗
  2. v1.6.0 AgentSwarms 1.6.0
    Release notes excerpt
    AgentSwarms 1.6.0 1.5 made the whole platform arrive at once. 1.6 adds a spreadsheet: Sheets, under Data & BI, with Excel's formulas over lakehouse tables. It adds a Settings page with themes and keyboard shortcuts. Most of the rest was found by using the product the way people do: edits that are no longer lost, failures that say so, runs that end the way they report, and sandboxes that no longer hold the lakehouse's credentials. 328 commits, 930 files, 9 migrations. Run npx supabase db push after pulling. --- Highlights Sheets: a spreadsheet over the lakehouse. …
    View on GitHub ↗
  3. v1.5.0 AgentSwarms 1.5.0
    Release notes excerpt
    AgentSwarms 1.5.0 1.4 put doors in the platform. 1.5 makes the whole thing arrive at once and then run itself: every service on every install, with nothing left opt-in; workflows as one graph over the four separate clocks that used to run the pipelines, the model builds, the notebooks and the swarms; a **transformation layer over the lakehouse; Delta Sharing** for people who are not users; SCIM so the directory creates and deactivates accounts; **column-level lineage recorded from what each run actually did; and a dashboard** that …
    View on GitHub ↗

FAQ

What infrastructure does the platform need?
The quickstart requires a Supabase project and Node.js 20.19 or later. The Docker setup starts multiple services; the documentation recommends about 8 GB of RAM for the full default install. Kubernetes deployment is also documented.
Does free software mean model calls are free too?
The repository uses the Elastic License 2.0. Model features require provider credentials, and call costs depend on the chosen provider and usage.
Can access to data and models be restricted before calls run?
The README describes server-side model rules, budgets, row filters and column masks, along with IAM and audit features.
Does the platform fail over automatically if its deployment region goes down?
No. The README says multi-region failover is not available; regional recovery relies on restoring from backup.
Is AgentSwarms Data and Agent Platform open source?
The source code is public on GitHub (AgentSwarms-fyi/agentswarms), but the repository does not declare a recognised licence, so the rights to reuse or modify it are unclear. Ask the author or check the README before building on it.
Is AgentSwarms Data and Agent Platform safe to use?
FollowAgents' FARS review (Nov 13, 2026) scored AgentSwarms Data and Agent Platform 21/29 on trust and safety. None of the eight safety controls the review checks for was missing outright. The repository has changed since that review.
Is AgentSwarms Data and Agent Platform still maintained?
Yes. The latest commit on GitHub is from 2026-11-20, within the last 30 days. The latest release is v1.7.0 (2026-11-14).
Who makes AgentSwarms Data and Agent Platform?
AgentSwarms Data and Agent Platform is published on GitHub by the account AgentSwarms-fyi. FollowAgents has not verified which person or company is behind that account.
What is AgentSwarms Data and Agent Platform's official website?
The project's GitHub repository lists https://agentswarms.fyi as its website. That link is set by the repository owner; the source code itself is on GitHub (AgentSwarms-fyi/agentswarms).
Was this page helpful?
View on GitHub ↗ Install ↓

Related concepts