Agentless IT automation
Automate the fleet, keep control of every run
FlowForge runs workflows across Linux and Windows servers over SSH and WinRM, with nothing installed on the hosts. Draft a workflow in plain language, watch it run live, and keep every action on the record.
100% agentless
Native SSH for Linux and network gear, WinRM for Windows. No agents to install or patch.
Drafted by AI
Describe the job in a sentence and get a workflow graph to test. Ask the AI to explain a failed step.
Runs you can see
Output streams live from every host, and sessions keep running if your browser closes.
Why FlowForge
Scripts and cron jobs hide what is happening and stop at the first failure.
Automate your runbooks without losing control.
FlowForge turns them into workflows you can see, pause for approval, retry and audit, from the first step to the last.
How it works
From servers to a running workflow
Add hosts one by one or sync them from your cloud inventory. A live handshake test checks the credentials before they are saved, and they are stored encrypted with AES-256-GCM.
Drag steps onto the canvas to form a graph from Start to End, or let the AI draft it. A pre-flight check flags circular dependencies and missing targets before anything runs.
Start it automatically, step by step through manual gates, or on a cron schedule. Output streams live to the browser, and every run is kept in the execution log.
AI assistant
Describe the job and the AI designs the steps, connects their dependencies, maps the commands for each operating system and puts the graph on the canvas for testing.
Example prompt: “Deploy update from git, compile, and restart nginx”
- 1Pull the update from git
- 2Compile and verify the build
- 3Restart nginx
Explain error On a failed run, the AI reads the stderr output and suggests a script to fix it.
Your choice of modelOpenAIAnthropicOllama, air-gapped
Administrators only. The example prompt is from the FlowForge page.
Drafted from a single sentence
Dashboard
The whole fleet, at a glance
The dashboard counts your workflows, servers and active, completed and failed runs. It streams the ten latest executions, each one a click away from its log, and shows administrators the worker pool against its hard cap of 50.
FlowForge · DashboardSample figures
- Workflows42
- Servers128
- Active18
- Completed1,204
- Failed7
Worker pool18 / 50
Latest executions
#882Git deploy and buildRunning#881Disaster-recovery drillCompleted#880Health check, 12 hostsCompleted#879Patch web tierFailed#878Restart Nginx and IISManual gate
Every run, live on the canvas
Steps light up as they run, finish, fail or wait at a manual gate for an operator. Pick how each step runs.
Workflow #882 Running
- 1Pre-flight checkValidate disk space and SSH connectivity across 12 target hosts.Completed
- 2Git deploy and buildPull the latest production branch, compile dependencies, verify binaries.Completed
- 3OpenSSH Windows scriptRun PowerShell to move Windows VMs from WinRM to OpenSSH.Running
- 4Restart Nginx and IISWaits for operator approval before restarting edge routing.Manual gate
Runs the script on the host, streams the logs live and ends on the exit code. For quick configuration scripts and health checks.
Long jobs run in a detached screen session and keep going if the browser closes or the connection drops. The engine polls for their status.
Opens both streams so an operator can answer prompts mid-run, with a 5-second settle so slow shell profiles finish loading first.
stdout · WebSocket
[direct] connected over SSH[direct] running health-check.sh[direct] disk 61% · load 0.42[direct] exit code 0[detached] screen -dmS nightly-build[detached] browser closed, job still running[detached] polling status… running[detached] finished, exit code 0[interactive] waiting 5s for the shell profile[interactive] Proceed with restart? [y/N][operator] y[interactive] nginx restarted- Running
- Completed
- Failed
- Manual gate
Access control
Three roles, hard boundaries
Role-based access is enforced in the interface and at the REST API. There is no self-registration: an administrator creates every account.
What the Super Admin can do
What the Administrator can do
What the Standard user can do
- DashboardFullFullAssigned only
- Run workflowsAllAllAssigned only
- Workflow canvasYesYesNo
- AI workflow generatorYesYesNo
- Servers and SSH keysYesYesNo
- Interactive terminalYesYesNo
- Audit logsAllAllOwn only
- Users and accessYesYesNo
Agentless execution
Nothing to install on the hosts: native SSH and WinRM from the FlowForge worker
Linux
SSH on port 22, to Bash or Zsh.
Windows
WinRM on port 5986, or OpenSSH: FlowForge writes the PowerShell script that switches a VM from WinRM to OpenSSH.
Network gear
Switches over SSH on port 22, from the same workflows.
Up to 50 workers run steps in parallel. Sample hosts.
Built for runs that must finish
Terminals that survive
The browser terminal runs through the FlowForge worker inside screen. Close the laptop lid and the session keeps running; vim, nano and htop work as they should.
Transfers that resume
Large files move in 4 MB chunks, each checked with SHA-256 and verified on arrival. Interrupted transfers pick up where they stopped.
Credentials sealed with AES-256-GCM, every action on the ledger
Passwords and .pem or .key files are encrypted at rest. The audit ledger records every execution, terminal command and configuration change, with the time, user, target, action and client IP.
- Immutable audit ledger
- Forward to Datadog or Splunk
- Users see only their own logs
| Time | User | Target | Action | Client IP |
|---|---|---|---|---|
| 02:00:04 | scheduler | web-01 | Run #882 started | 10.0.4.12 |
| 02:07:31 | ops.lead | web-01 | Approved gate: Restart Nginx | 10.0.8.21 |
| 02:09:12 | ops.lead | db-01 | Terminal: systemctl status | 10.0.8.21 |
| 09:15:40 | it.admin | FlowForge | Role changed: Standard user | 10.0.2.7 |
In the box
Everything around the run
Pre-approved runbooks
Run approved blueprints, such as a disaster-recovery drill, with one click.
Workflows from your repositories
Connect your GitHub organisations to pull runbooks, scripts and workflow YAML on every commit, or import .yml pipelines and run them on your own infrastructure.
Reliability over time
Track workflow reliability, SLA compliance and throughput across the fleet.
Told when it matters
Send SLA breaches and failed runs to webhooks, Slack or Microsoft Teams.
Only what you allow
Switch whole modules on or off across the platform to match your own rules.
Where it fits
End-of-day batch work, without the 2 a.m. surprises
Banks and other regulated teams run reconciliation, reporting and ledger updates between close and next-day open. These are the six things that work needs, and how FlowForge covers each.
Read the EOD automation playbook- 01
No manual dependency
Workflows start on a schedule or a trigger, with no one at the keyboard.
- 02
Real-time visibility
Logs, status and timing for every job, as it happens.
- 03
Interactive automation
Operators approve at gates or answer prompts without breaking the run.
- 04
Crash-resilient runs
Detached sessions keep going when a browser or connection drops.
- 05
Granular control
The graph shows which steps depend on which, so the right branch is the one you deal with.
- 06
Full audit trail
Every human and system action, logged with who, what and where.
Questions, answered
Read the documentation
Do I need to install an agent on my servers?
No. FlowForge connects over native SSH for Linux and network devices and WinRM for Windows, so there is no agent to install or patch.
Which AI models can it use?
An administrator connects the model in AI settings: OpenAI, Anthropic, or a local Ollama deployment for air-gapped sites.
What happens if my browser closes during a run?
Detached steps and terminal sessions keep running inside screen on the worker. Reconnect and pick up where you left off.
Can people sign themselves up?
No. Enrolment is closed: an administrator creates each account and assigns its role, and standard users see only the workflows mapped to them.
Does it work with our directory?
Yes. Sign-in binds to Active Directory or LDAP, then checks for a local FlowForge profile before granting access.
Can we send its logs to our SIEM?
Yes. Telemetry, sign-in events and workflow logs can be forwarded to platforms such as Datadog or Splunk.
See it on your fleet
Bring your runbook
Tell us what you run today and we will show you how it looks in FlowForge.