Hermes
An open-source agent harness for businesses that need model choice and greater control over infrastructure. Open source does not mean free or secure by default.
CUSTOM IMPLEMENTATION / THREE MONTHS
That process can move forward without someone chasing it all day.
Show us how it works today, which information it needs and where it stalls. We build an agent that consults authorized sources, prepares the work and uses the agreed tools. Your team knows what it can do, what needs approval and what happens when information is missing.
3 monthsto build and test
Process firstarchitecture follows the work
Human reviewfor important decisions
USD 5,000starting point
WHEN A MANUAL PRACTICE IS NO LONGER ENOUGH
The implementation connects part of the work to the tools your business already uses.
Applied AI training helps a team document and run a task step by step. An implementation begins when an agent needs to consult sources, move information between tools, work on a schedule or prepare actions for approval. The training program is not a prerequisite when the process and responsible person are already clear.
BEFORE CHOOSING A PLATFORM
An agent needs context, tools, permissions and supervision.
Context
Approved sources, examples, rules and memory so the agent does not start from zero.
Tools
Explicit connections for reading, preparing or executing the contracted work.
Permissions
Limits on what the agent can access, change or send.
Supervision
Logs, alerts and a responsible person for exceptions and high-risk decisions.
THE TECHNOLOGY FOLLOWS THE WORK
The agent's responsibility determines the right foundation.
Codex
An OpenAI-based environment suited to software, files, skills, plugins and connected services. Access and limits depend on the company's plan and workspace.
Both
A combined architecture only when each agent has a clear responsibility and the handoff can be tested and reviewed.
WHERE AUTOMATION CAN START
We review the business and automate only work with rules, sources and a reviewer.
Possible areas include leadership and operations, sales, marketing, customer service, administration and finance, HR, procurement, inventory, projects, legal, data, software, design and documents. The assessment opens possibilities; the proposal states exactly what is included.
A person repeats the same work every day
We identify what arrives, which decisions rely on memory and what must be left ready.
Information lives in too many places
We determine which source is authoritative and who has permission to use it.
The team uses ChatGPT but keeps starting over
We preserve instructions, files and decisions in an operable context.
A previous automation broke
We examine the normal case, the exception and the person who needs to step in.
YOUR TEAM KNOWS THE WORK / WE BUILD THE AGENT
We turn operating experience into rules an agent can follow.
Your business brings
Responsible people, good and bad examples, authorized sources, permissions, exceptions and a human review standard.
UNSU builds
The architecture, environment, context, connections, agents, skills, memory, tests, alerts, documentation and handoff stated in the proposal.
BUILT WITH REAL CASES
We understand first, connect second and try to break it before handoff.
MONTH 01
Understand and choose
Map inputs, sources, rules, exceptions, risks, owners and technical scope. Then choose Hermes, Codex or a justified combination.
MONTH 02
Build and connect
Configure the environment, prepare context and connect the tools included in the proposal. Initial cases run under supervision.
MONTH 03
Test and hand off
Run incomplete inputs, unusual formats and real exceptions. Adjust rules, permissions and alerts, then train the operating team.
AUTOMATION IS NOT ABANDONMENT
The agent performs the agreed work. Your business keeps the important decisions.
Every workflow defines what the agent may consult, what it may change, what it must only prepare and what requires approval. Financial, legal, employment and other high-risk decisions retain a responsible person.
WHAT STAYS WITH YOUR BUSINESS
This is an operable implementation, not a demo.
Automation map
Reviewed processes, priorities, sources, owners, risks and later opportunities.
Architecture decision
Hermes, Codex or both, with rationale and dependencies.
Configured environment
AWS, local infrastructure or another agreed environment with access and costs documented.
Agents and workflows
The context, skills, memory, schedules, connections and actions expressly included in scope.
Tests and controls
Normal cases, exceptions, logs, alerts, permissions and human checkpoints.
Runbook and training
Instructions to operate, review and modify the implementation, followed by handoff to the responsible people.
Ownership: your data, accounts and outputs remain yours. The proposal identifies the configuration and code you receive, open-source components and required external services.
CUSTOM PROGRAM / THREE MONTHS
From USD 5,000
Includes diagnosis, architecture, configuration of Hermes, Codex or both, contracted workflows, testing, documentation and handoff. Final price depends on the number of processes, tools, sources, integrations and security requirements.
AWS, models, APIs, databases, connectors and licenses are billed separately when needed. The proposal states those costs, the payment schedule and which processes or departments are included before work begins.
Discuss my implementation ↗COMMON QUESTIONS
What businesses usually ask before we build.
Do I need to choose Hermes or Codex first?
No. We first inspect the process, existing accounts, permissions and operating environment. The architecture is part of our work.
Will you automate every department?
No. We map the areas you care about, while the proposal names the processes included in the three-month scope.
What about sensitive information?
We work only with authorized sources and access. Special security or compliance requirements are reviewed before scope because they can change architecture and price.
What happens after three months?
Your team receives the contracted implementation and documentation. Ongoing support, monitoring or additional processes are scoped separately.