loader
Redeyed Agents Autonomous • First-party

Agents that do the work — while you watch

Set a goal. A Redeyed agent plans it, uses every tool in its registry, runs code in a sandbox, and delivers inside Laboratory—every step visible in real time on infrastructure Redeyed operates. Approve anything sensitive; it routes around what you decline.

How an agent runs

Goal in. Result out. Nothing hidden.

Every agent runs the same transparent loop — and you see each turn as it happens.

GoalYou describe what you want.
PlanThe agent breaks it into steps.
ActUses tools & runs code.
VerifyChecks its own work.
DeliverHands you the result.
Why Redeyed Agents

Capable, transparent, and entirely yours

The autonomy of a top coding assistant — running on your own models and hardware, with you in control.

Autonomous loop

Agents keep planning, acting and self-correcting until the goal is met — no babysitting each step.

26 built-in tools

Research, security intelligence, code, content and orchestration — the agent picks the right one for the task. See the catalogue.

Sandboxed code

Agents run real code in an isolated Docker sandbox, so they can build and test — safely contained.

You stay in control

Approve sensitive actions before they run, or decline them — the agent simply routes around what you refuse.

Scheduled & recurring

Give an agent a cron schedule and a standing goal — it runs on its own and reports back.

Teams that delegate

An orchestrator can spin up sub-agents, split the work, and bring the pieces back together.

What an agent can actually do

26 tools, one decision loop

An agent picks the tool the goal needs. Every one of these is live today — nothing here is a roadmap item.

Everything an agent can reach for.

Fetch page

Pull a public URL and read its text, tags stripped.

fetch_url

Knowledge search

Recall notes saved from earlier runs and conversations.

knowledge_search

Knowledge write

Save a verified fact for later recall.

knowledge_write

Summarize

Condense a block of text, to your instructions.

summarize

Web search

Search the live web and return ranked results with snippets.

web_search

Breach search

Look an address up in Redeyed's breach intelligence.

breach_search

IP reputation

Score any public address against Sentinel's reputation graph.

ip_reputation

Recon scan

Passive DNS, SPF/DMARC/MX posture and reputation for a domain.

recon_scan Non-intrusive

Threat intel

Check an IP, domain, URL or file hash for risk signals.

threat_intel

HTTP probe

Send a raw request to the authorized target and read the response.

pentest_http Verified target only

Headless browser

Render an in-scope URL and return DOM plus console output.

pentest_browser Verified target only

Origin discovery

Find the real origin behind a CDN or WAF.

origin_discovery Verified target only

Record finding

File a validated finding with severity and evidence.

record_finding Verified target only

Security tooling

Run scanners in the sandbox against the authorized target.

pentest_exec Verified target only

Git

Status, diff, commit and log inside the sandbox.

git Sandboxed

List files

Walk the workspace as a depth-limited tree.

list_files Sandboxed

Read file

Read a file from the sandbox workspace.

read_file Sandboxed

Run tests

Run the project suite, auto-detecting the runner.

run_tests Sandboxed

Shell

Run a command in the isolated sandbox for this run.

shell Sandboxed

Write file

Create or overwrite a file in the workspace.

write_file Sandboxed

Faceless video

Script, illustrate, voice and assemble a short from a topic.

faceless_video

Generate clips

Cut short vertical clips out of a long source video.

generate_clips

Generate image

Create an image from a prompt and attach it to the run.

generate_image

Text to speech

Render a spoken voiceover from text.

text_to_speech

Send email

Send a plain-text email to one recipient.

send_email Needs your approval

Delegate

Hand a sub-task to a member agent and take back its result.

delegate

Showing all 26 tools

Start from a shape

Six presets, or build your own

A preset is just a starting allow-list. Add or remove any tool afterwards — an agent can only ever use what you granted it.

Researcher

5 tools

Reads the web, keeps what it learns, and reports back.

  • Web search
  • Fetch page
  • Summarize
  • Knowledge search
  • Knowledge write

Recon

5 tools

Passive exposure and reputation work on a domain or address.

  • Recon scan
  • IP reputation
  • Threat intel
  • Breach search
  • Summarize

Pentester

8 tools

Actively tests a target you have verified you own, and files what it validates.

  • HTTP probe
  • Security tooling
  • Headless browser
  • Origin discovery
  • Record finding
  • Fetch page
  • Recon scan
  • Summarize

Recon Pro or Enterprise

Developer

6 tools

Reads, writes and tests code in a container it cannot escape.

  • Shell
  • Read file
  • Write file
  • List files
  • Run tests
  • Git

Creator

5 tools

Turns a topic into images, voice and short-form video.

  • Web search
  • Generate image
  • Generate clips
  • Faceless video
  • Text to speech

Orchestrator

4 tools

Splits a job across member agents and assembles the answer.

  • Delegate
  • Web search
  • Summarize
  • Knowledge search

Need a tool that isn't here?

Tell us what the agent would need to reach. We add tools to the registry regularly, and the ones customers ask for go first.

Registry 26 Tools available

Across 7 categories — research, security, offensive testing, code, content, comms and orchestration.

Control 3 Autonomy levels

Suggest, ask, or auto. Coding, comms and spend stay approval-gated whichever you pick.

Bounded 20 Step ceiling by default

A run that stops making progress is stopped rather than left to burn budget in a loop.

Get started

Three steps to your first agent

From a plain-language goal to a finished result.

Describe the goal

Pick a preset or write your own — “research X”, “fix the failing test”, “turn this into a short”.

Watch it work

The run streams live: its thinking, each tool call, and every result — just like the panel above.

Approve & collect

Green-light anything sensitive, then collect the artifacts the agent produced. Re-run any time.

Build with agents

Drive agents from your code

Launch and monitor the fleet with the official SDKs — or run agents right in the Lab. No setup beyond your API key.

redeyed — install
$
Open source

SDKs & apps on GitHub

Everything you need to build on Redeyed — MIT-licensed, free to use.

github.com/Bruted

Put an agent to work today

Deploy your first agent free, watch it work end to end, and keep everything it makes — all inside Redeyed.