Find the answer
Explore a question, compare sources and turn scattered information into a recommendation you can act on.
Self-hosted agents for your team
A complete
agent platform
Under your control
Give your team agents that research, write, code and work across your systems. Start with a complete platform, then choose the models, tools and infrastructure that fit.
The alternative to managed AI workspaces
Yours to adapt
01 / THE ROOFMANAGED AGENTS
Put an agent
on the job
An AI agent can take a task from a goal to a result. It uses your knowledge and tools to work through the steps, make decisions and take action.
Move work forward
Turn your team's context into a brief, a report or a code change, ready for review.
Make room for your team
Hand over recurring tasks. Keep information up to date, route incoming work and flag what needs your attention.
AN AGENT IN PRACTICERESEARCH EXAMPLE
YOUR ASSIGNMENT
“Find and compare suppliers, recommend the best fit for our next project.”
AGENT AT WORK
- Find suppliers that match the project brief
- Compare capabilities, costs and trade-offs
- Check the evidence behind the shortlist
READY FOR YOU
A recommendation, with reasons
A comparison brief with a shortlist, trade-offs and links to the sources.
You set the direction. Follow progress and step in when it matters.
02 / THE PILLARSEXPERT · WORKSPACE · CONNECTORS
A few powerful parts
A whole world of agents
Choose an expert, a workspace and the connections it needs. A job brings those saved parts together to do the work.
Researcher configuration01 02 03
expert.yaml
THE EXPERT
Model, instructions and tools
kind: Expert
metadata:
name: supplier-researcher
spec:
harness: srw
model: local/gemma4
settings:
temperature: 0.2
systemPrompt: |
Find suppliers. Compare the evidence.
Cite sources. Explain the trade-offs.
tools:
- web.search
- web.fetch
- files.read
- files.write
- knowledge.search
- database.queryworkspace.yaml
THE ENVIRONMENT
A file workspace, backed by S3
kind: WorkspaceTemplate
metadata:
name: research-s3
spec:
backend: virtual
storage:
provider: s3
bucket: team-workspaces
prefix: research/
persistence: projectjob.yaml
THE JOB + CONNECTORS
The task and the systems it needs
kind: Job
metadata:
name: compare-suppliers
spec:
task: |
Find suppliers. Recommend the best fit.
expert: supplier-researcher
workspace: research-s3
tags: [local, procurement]
connectors:
knowledge: purchasing-guide
cloud: team-projects
database: supplier-catalog
output: supplier-shortlist.mdExpertWorkspaceConnectors
The job selects saved parts by name. All three examples share the team-projects folder.
FROM DEFINITION TO WORK
Explore an agent example Follow an agent from assignment to result
03 / THE FOUNDATIONTHE SRW ORCHESTRATOR
You define the agent
SRW runs the system
Your manifest is the starting point. SRW handles deployment and coordinates the agent through the work, on your infrastructure.
- 01 / PREPARE
From manifest to ready
Resolve the definitions, prepare the workspace and connect the systems the agent needs
- 02 / RUN
Put the agent to work
Dispatch the work, monitor execution and preserve its context through pauses and continuation
- 03 / DELIVER
Bring the work back
Collect results and the record of the work, ready for your team to review and use
SRW OrchestratorDeploy · Coordinate · Monitor
THROUGHOUT THE WORK
You stay in the loop
Progress and questions come to you
Give direction, pause work and review the result.
YOUR NEXT STEP
See SRW
On your terms
Get notified when hosted SRW launches, or talk to us about a pilot for your team.
PLANNED LAUNCH / WINTER 2026–27Have a project in mind?
Tell us what you’d like an agent to handle. The first call is free. We’re currently taking on pilot projects: about a month, set up together with your team.
Discuss a pilot sales@superhuman-remote-worker.comHOSTED SRW
Get notified
Want SRW hosted for you? We’ll email you when it launches.
Source available — explore the code