The AuxForge for shipping production software

Specialized agents for code, terminal, APIs, databases, and QA—coordinated by an orchestrator that shares context across your entire repo. Ship complex changes with guardrails, previews, and automation built in.

One workspace, many specialists

A multi-agent system that routes work to the right tool—Editor, Terminal, API, DB, and QA—while sharing project-wide context and audit trails for every action.

🎯

Smart tool routing

Orchestrator decides which agent should act and passes the right context.

🧠

Project memory

Index your repo (code, docs, OpenAPI, manifests) with AST search and embeddings.

🛡️

Guardrails

Permissions, human-in-the-loop approvals, and full audit logs.

Example tasks

  • "Add OAuth login end-to-end" (plan → routes → deps → env → UI → tests → preview → validate)
  • "Refactor payment code + add retries" (repo plan → targeted edits → unit tests → log verify)

Premium AI Code Editor + File Manager

Get multi-file diffs, rollback, and PR-ready summaries. Transform entire repos with refactors, codemods, and test generation—then gate merges on passing tests.

Premium agent
📝

Structured edits, repo-wide impact

  • PR-grade changes: Multi-file patches, diffs, explanations, revert/rollback
  • Big moves: Refactors, codemods, dependency upgrades, dead-code cleanup
  • Test-aware: Propose and run tests automatically after changes
  • Deep context: Pulls in relevant files, symbols, stack traces, and failures
  • Git-native: Branching, commits, PR creation, commit message generation

Example tasks

  • "Migrate from React Router v5 to v6 across the repo with tests."
  • "Increase test coverage to 85%; generate missing tests."

Terminal Agent

Safe, adaptive command execution. Run installs, builds, migrations, and tests with allow/deny rules. The agent reads output, fixes issues, and retries safely.

  • Guarded commands: Configurable allow/deny rules for risky ops
  • Output-aware: Reacts to errors (e.g., missing peer deps, TS compilation)
  • Artifacts: Preserve logs and cache builds for diffable runs

Example tasks

  • "Set up Playwright, record first test, run it in CI mode."
  • "Run db migration; if it fails, suggest a fix and retry."
🌐

API Agent (Postman-like)

API client + AI test generation. Build and test APIs faster with collections, auth, OpenAPI import, and automated monitors.

  • Request composer: Collections, environments, auth helpers, variables, scripts
  • OpenAPI in: Generate collections, typed clients, and contract tests
  • Automated tests: Happy-path and negative tests, schema checks, monitors/schedules
  • Trace-aware: Correlate with server logs; capture headers and bodies

Example tasks

  • "Import Stripe spec, generate a typed client, and create a smoke suite."
  • "Nightly monitor for health endpoints; alert in Slack on failures."
🗄️

Database Agent

Query, explore, and migrate safely. Explore schemas, generate queries, visualize results, and run migrations with explicit approvals and rollbacks.

  • DB explorer: Schemas, relations, indexes, ER-style views
  • Query assistant: Safe read-only by default, query generation, explain plans, exports
  • Migration assistant: Generates SQL migrations + rollbacks, staged runs with approvals

Example tasks

  • "Find top 20 slowest queries from last hour; propose indexes."
  • "Add column with default; backfill and create migration + rollback."
👁️

Preview + Vision/QA Agent

See it, test it, ship it. Spin up live previews, capture screenshots, validate against designs, and generate UI/API automations with artifacts.

  • Live preview: Run dev servers; capture screenshots on demand
  • Visual checks: Detect layout regressions, contrast issues, missing elements
  • UI automation: Generate Playwright/Cypress tests from user stories
  • API automation: Contract and scenario tests with status/schema/side effects

Example tasks

  • "Implement new header; match Figma ≥95%; open diffs if not."
  • "Create an end-to-end checkout test; attach screenshots and videos."
🔒

Collaboration, Observability, Security

Built for teams and compliance. Share sessions, review diffs, and keep a complete audit trail across agents—inside sandboxed, ephemeral workspaces.

  • Real-time reviews: Comment on diffs, commands, API calls, and queries
  • Audit trail: Every action logged—actor/agent, inputs, outputs, diffs, artifacts
  • Secure by default: Ephemeral containers, secrets management, RBAC

Cost-optimized model strategy

Premium where it matters, efficient everywhere else. The Editor runs on a top-tier model. All other agents use fast, hosted OSS models with smart escalation only when needed.

💎

Premium Editor

GPT-4o/Claude/Gemini-class by default; BYOK supported.

Hosted OSS

Llama, Qwen, Mistral variants via vLLM/TGI for low cost and speed.

🎯

Smart Router

Escalate to premium only for complex refactors and hard cases.

BYOK available to reduce costs and keep your model choices flexible.

Pricing that scales with your workflow

Low-cost agents feel unlimited. Pay for premium code edits and reliable automation when it matters. BYOK supported.

Free (Hobby)

$0
Individual tinkering and evaluation
  • Low-cost agents: fair-use pooled access
  • Editor agent: 100 edit actions/month
  • 1 project, 1 preview environment
  • 30 minutes/month preview/test compute
  • Community support
Get started free

Team (Startup)

$39
per month per user
  • Editor agent: 2,000 edit actions/user/month
  • Low-cost agents: priority with concurrency
  • 20 projects, 20 preview environments
  • 20 hours/user/month compute (pooled)
  • 1,000 UI/API automation runs/month
  • Git PR integration, branch protections
  • Team collaboration and RBAC
  • Audit retention: 30 days
Start Team

Enterprise

Custom
Contact us for pricing
  • Custom model routing; VPC/on-prem options
  • SSO/SAML, SOC2, IP allowlisting
  • Unlimited projects/environments
  • Higher SLAs, priority support
  • Extended audit retention (90-365 days)
  • SIEM export
Talk to sales

How we stack up

The only AuxForge with first-class API and DB agents, live preview with vision-based checks, and safe multi-agent orchestration.

Feature Us Cursor Bolt.new Windsurf Lovable
Multi-agent specialization (Terminal, API, DB, QA) Partial Partial
AI code edits with repo-wide diffs and PRs Partial Partial
Built-in Terminal with agent control and guardrails Partial Partial
Postman-like API client with AI test generation
Database explorer + AI queries + migration assistant
Live preview with screenshots and vision checks Partial Partial
Vision-based UI validation against designs/specs
Orchestrated tool routing with shared context Partial Partial
Cost-optimized model strategy (OSS + Premium) Unknown Unknown Unknown Unknown
Deep audit trail (commands, queries, tests, diffs) Partial Partial Partial Partial

What sets us apart

🔧

Complete toolkit

First-class API client and DB explorer, not just an editor

👁️

Visual validation

Live preview with screenshot-based UI validation loops

🎯

True orchestration

Multi-agent coordination with guardrails and approvals

🤖

Built-in automation

Generation and execution of UI and API automation

💰

Cost optimized

Premium only for the editor, OSS everywhere else

Plug into your stack

First-class integrations and a plugin SDK to bring your internal tools and connectors into the workbench.

🔗

Git

GitHub, GitLab, Bitbucket

🚀

Deploy

Vercel, Netlify, Fly, Render

💬

Chat

Slack, Teams

🔐

Secrets

Vault, Cloud KMS

⚙️

SDK

Custom CLIs, internal APIs, private DB connectors, CI runners

Frequently Asked Questions

Can I bring my own API keys?

Yes—BYOK is supported for the Editor (and optionally others).

Are low-cost agents unlimited?

They're fair-use with rate limits, not token-metered.

What's an "edit action"?

A successful, applied multi-file change with diffs.

Is this secure?

Yes—ephemeral sandboxes, secrets management, RBAC, and full audit trails.

How do you meter usage?

  • Editor action: One accepted multi-file change request (plan → patch → diff). Only counted when changes are applied.
  • Automation run: One UI or API test execution, including screenshots/videos/artifacts.
  • Compute minute: Active dev server, preview, or headless test execution minute in your sandbox.
  • Low-cost agents: Not token-metered. Fair-use via rate limits and burst controls.