MCP Hub & Integrations
An MCP marketplace agents plug into on demand — ticketing, comms, docs, CI/CD, monitoring & design — with heavily-used servers promoted to native CLIs for lower token cost.
Self-hosted · your code never leaves your network
Ship the whole job, not the keystroke.
automatedev.ai is the agentic OS for software engineering. One idea goes in — a tested, reviewed, merged pull request comes out. It plans, builds, tests, reviews and secures your software on its own, with the engineer in control at every gate.
Runs in your wallszero data exit
Self-repairbuild · test · git retry
The live pipeline
One idea. One autonomous run. One pull request.
The OS owns the entire loop. Every phase has explicit success criteria, runs in isolation, and is observable end-to-end — control flow lives in code, never a fragile prompt-chain.
1
Understand
IntakeYour idea and files become a clear, written brief.
2
Plan
PlanA short back-and-forth writes and locks the plan at a gate.
3
Build
BuildAn isolated run per repo — own workspace & database.
4
Verify
CheckTested end to end, reviewed, and scanned for security — findings loop back as fixes.
5
Ship
ShipA finished change as one pull request, authored as you.
The loop closes on itself — every reported defect becomes a run the OS picks up, reproduces, fixes, and ships.
1Idea in → PR out
∞Parallel runs, fully isolated
5Loop phases, end to end
2Cloud — or your own servers
Plugs into your stack
Connected to the tools your team already runs.
Your AI engine of choice, GitHub & GitLab, Vault for secrets, Playwright, Postgres & Docker — wired today, with more on the roadmap.
CClaude CodeAI provider · default
CCopilotAI provider
CCodexAI provider
OOpenCodeAI provider
KKimiAI provider
AAntigravityAI provider
KBYOKYour own API key
CClaude CodeAI provider · default
CCopilotAI provider
CCodexAI provider
OOpenCodeAI provider
KKimiAI provider
AAntigravityAI provider
GGitHubSource
GGitLabSource
VVaultSecrets
PPlaywrightTesting
PPostgresData
DDockerDeploy
GGitHubSource
GGitLabSource
VVaultSecrets
PPlaywrightTesting
PPostgresData
DDockerDeploy
The actual product
This is the console your team logs into.
Not a concept render — the real self-hosted control plane. Ideas, defects, implementation, review and testing live in one place — each an autonomous run you can watch and gate.
Everything, in one OS
A complete engineering platform, not a copilot.
From idea intake to autonomous debugging, every capability runs on the same deterministic control plane — a shared brain, isolated runs, and a human gate only when judgment is needed.
01
One approval, every phase
Approve the plan and a single session drives implement → test → review → secure → PR across every repo — retrying build, test and git failures on its own, and resuming mid-pipeline after any restart.
02
Every run in a sealed sandbox
Each run gets its own workspace, network sandbox and throwaway database — so many versions of the same app run at once, on the same ports, with zero collisions.
03
Tests the real app, shows its work
A fresh agent drives the running app through acceptance scenarios and attaches screenshot proof. Pass opens the PR; fail re-implements with the failure in hand.
04
Reviewed & hardened before merge
An autonomous reviewer checks correctness, style and reuse; a dedicated security pass scans for injection, access and secret-handling flaws — every finding routes back into the fix loop.
05
Bring any model, or your own key
Claude Code, Copilot, Codex and more install on demand behind one factory — or bring your own API key. Swap the engine without touching the pipeline.
06
Watch it work, gate every action
A live stream of the agent’s text, thinking, tool calls and file diffs — with allow, always-allow or reject on any action, and a stop button that halts a run mid-flight.
Everything else, in the box
Source control — GitHub & GitLab
One-command self-hosted deploy
Secrets in Vault, never the database
Autonomous conflict resolution
Skills hub, materialised into runs
Teams, projects & role-based access
Attention inbox & live agent panel
Command palette & realtime events
Autonomous debugging from a defect
Per-run usage & cost telemetry
Token-cost compression
Crash-resume mid-pipeline
Shared knowledge
One shared memory, so the team and the agents never lose context.
-
A living map of your code
Built from your own code and docs and kept current — a map of how everything fits together, across every repo and team.
-
Remembers every decision
Why things were built the way they were is captured and kept. Every shipped feature enters the memory — one source of truth.
-
Feeds context back in
The right context is handed to each run up front, so agents work from what’s already known instead of re-reading everything.
Cloud
Reach
Hosted SaaS, desktop & web apps for instant adoption.
Self-hosted
Walls
Source & secrets never leave your network.
Runs anywhere
Cloud for reach. Inside your own walls for compliance.
Regulated and IP-sensitive teams can't send source to a third party — so they run automatedev entirely inside their network. Same OS, same loop, zero data exit. It's the thing cloud-only rivals can't offer, and the reason enterprise pays.
Deterministic orchestration
Control flow lives in code — an event bus, workflow engine & gate manager that outlast any single model.
Whole job, not the keystroke
From idea to shipped PR, the OS owns the engineering workflow around the code — not just the typing.
On the roadmap
The vision beyond today's core.
These capabilities aren't shipped yet — here's where automatedev is heading next. Everything above runs today; everything below is planned.
Mobile Build & Test
Mobile apps build in the same pipeline, with Appium & device test CLIs mirroring the Playwright tester — end-to-end automation that extends to mobile, not just web.
Production Self-Healing
Continuous log & APM monitoring across dev and production that auto-detects incidents, builds a reproduce pack, then fixes, tests & opens a PR — closing the loop from live failures.
Business PRD Layer
A stakeholder PRD step for business & PM users, brain-informed about code intricacies, UI themes & constraints — flagging the docs, APIs & approvals a feature needs up front.
Custom Workflow Builder
Chat to build a workflow, generated from your plan & prompts. Tag it to an agent and a trigger point, then run it on a schedule or an event.
Offline License Gating
A pure offline, signed license key for air-gapped self-hosted deployments — entitlement enforced without ever phoning home.
Planned integrations
JJiraTicketing
LLinearTicketing
SSlackComms
TMS TeamsComms
CConfluenceDocs · RAG
JJenkinsCI / CD
VVercelCI / CD
NNew RelicMonitoring
EELK / KibanaMonitoring
FFigmaDesign
Pricing
Run it in the cloud, or entirely in your walls.
Two ways to adopt the same platform. Talk to us and we’ll scope the right one for your team.
Teams · Cloud
For the squad
Hosted by us- The full idea → merged-PR pipeline
- Shared projects & an approval layer
- Review, security & debug loops
- Shared knowledge, skills & role-based access
Self-hosted
Enterprise · In your walls
For the org
Talk to us- Runs entirely on your own servers
- Your code & secrets never leave your network
- Teams, governance & role-based access
- One-command deploy, plus onboarding & support
FAQ
Questions, answered.
What automatedev.ai is, how it runs, and how it stays under your control.
What is an agentic OS for software engineering?
It is a control plane that runs AI coding agents through the full engineering job — not just autocomplete. You give automatedev.ai an idea, a defect, or a ticket, and it plans, writes, tests, reviews, and secures the change as one autonomous run, ending in a merged pull request. A human engineer approves at every gate.
Can automatedev.ai run fully self-hosted?
Yes. automatedev.ai is self-hosted and single-tenant: one deployment per client, running entirely on the client's own infrastructure. Your source code never leaves your network, so there is zero data exit. A hosted cloud option is also available for teams that prefer not to run their own infrastructure.
Which AI models or coding CLIs does it use?
automatedev.ai uses a vendor-neutral provider factory. Claude Code is the reference and default engine. GitHub Copilot, Codex, Opencode, Antigravity, and Kimi install on demand, and you can bring your own API key. You can swap the underlying engine without changing the pipeline.
How does it keep the engineer in control?
Every run streams the agent's text, thinking, tool calls, and file diffs live. You can allow, always-allow, or reject any action, and a stop button halts the run instantly. Autonomy is gated at checkpoints — the PRD, the plan, and pre-merge or release — so nothing ships without human approval.
Which tools does automatedev.ai integrate with?
It integrates with GitHub for source control and Jira for ticketing. Secrets use a pluggable provider with a HashiCorp Vault adapter and a built-in encrypted store. Every integration sits behind a swappable adapter interface.
What does automatedev.ai cost?
Pricing is quote-based with two ways to adopt the same platform: Teams · Cloud, hosted by automatedev.ai, and Enterprise · In your walls, fully self-hosted and single-tenant. Contact hello@automatedev.ai to scope the right option for your team.
What engineering workflows does it support?
Feature development, debugging, review and refactor, and release are all first-class workflows. An onboarding workflow maps your codebase first so the others run with full context.
How is automatedev.ai different from GitHub Copilot or Cursor?
Copilot and Cursor accelerate the keystroke — they help an engineer write code faster in the editor. automatedev.ai owns the whole job: it runs the plan-build-test-review-secure pipeline autonomously and returns a reviewed, tested, merged pull request. It is self-hosted and single-tenant, and it orchestrates coding agents rather than being one.
See it turn one idea into a merged PR.
Book a 30-minute demo and watch a real run go from brief to reviewed, tested, merged pull request — in the cloud or entirely inside your own walls.