NeyruOne personal identity.
A verified wallet returns to the same Neyru agent. The workspace organizes your projects, conversations, files, and saved preferences around that identity.
GET TO KNOW NEYRU
A clear place
for complex work.
Bring the goal, the files, and the context. Your personal agent uses the configured tools to move the work forward, with the process and results in one place.
YOUR WORKSPACE
Your personal agentContext that belongs together.
- Projects
- Reference files
- Saved instructions
FROM A GOAL TO A RESULTIllustrative flow
- 01
Define the work.
Describe the outcome. Add the constraints and material that matter.
YOUR GOAL + CONTEXT - 02
Use the right tools.
Hermes coordinates the allowed models and tools on a configured deployment.
AGENT + CAPABILITIES - 03
Inspect the result.
Review the response, evidence, and files. Continue with a follow-up task.
RESULT + NEXT STEP
This diagram explains the workflow. Live task execution depends on the connected backend and verified tools.
AvailabilityThe right tools for the work.
Hermes provides the agent runtime. Configured model routes and tools support different tasks, with the project inference credential kept on the server.
Results you can inspect.
Task history, progress events, and generated files make the work easier to review. Good output still needs evidence, checks, and a clear account of uncertainty.
A generalist, with practical tools
Start with the work.
Choose the useful result.
These describe the configured product's intended workflows. Live capability depends on the engine, model routes, and tools enabled by the operator.
Research
Make a question easier to act on.
Read provided public pages, use configured search, and bring findings into a comparison or written brief. Ask for sources and have Neyru distinguish evidence from its interpretation.
A useful resultA sourced brief or comparison
Search requires its own configured service. Public-page fetching does not include interactive browser control.
Build & code
Turn a brief into something testable.
Use a configured coding route and the Hermes sandbox to write or inspect code, work with files, and report the checks that actually ran. Keep setup instructions beside the output.
A useful resultSource files with usage notes
Execution requires the validated sandbox. Generated applications are not automatically deployed or hosted.
Data & files
Find a clearer story in the material.
Work from supplied reference files, inspect data, and use code where calculation or transformation helps. Ask for assumptions, missing values, and checks alongside the findings.
A useful resultAnalysis notes and output files
Supported formats depend on the configured tools. Exact calculations and file integrity need verification.
Vision
Give an image a useful second look.
Pass a supplied image to a vision-capable route for descriptions, observations, or feedback. Keep visible details separate from assumptions about what the image does not show.
A useful resultImage analysis or concrete feedback
Vision is a model capability. It must be configured and tested separately from text-only routes.
Images
Bring a visual direction into the work.
Image generation and editing have dedicated tool adapters. A configured service can take a brief or reference and return an image file to the workspace.
A useful resultAn image output to review
These routes require their own verified endpoint, model capability, and pricing. Live generation and editing have not yet been tested.
Repeatable work
Give recurring work a place to return.
Set up interval-based tasks tied to a project and review each run in the workspace. Each execution uses the available runtime, context, and task allowance.
A useful resultA record of each scheduled task
Schedules need the server-side worker to stay running. Arbitrary sandbox processes do not persist between completed tasks.
Context, with a little control
Keep the agent close.
Keep the controls closer.
A personal agent should be understandable. The workspace gives the work a home and makes the important controls visible.
Read the user guideProjects separate the context
Group related conversations and files. Project notes are scoped to the project, while wallet-wide preferences are available across your workspace.
Memory stays editable
Add, review, edit, and remove explicit memory entries. Current memory is user-managed; automatic global learning is disabled.
Progress and Stop are part of the task
Review events, task status, output, and error messages. A Stop request is handled by the server-side worker and engine.
Boundaries belong in the product
Tasks have spending limits and a reserved-use ledger. Unknown provider charges remain held for review rather than being reported as zero.
Current product boundaries
Know what you're looking at.
The source implementation, an interface preview, and a verified live agent service are different stages. Here is the current boundary.
| Area | Included in the source | Current limitation |
|---|---|---|
| Workspace | Projects, files, conversations, manual memory, schedules, and task views. | The hosted static preview cannot save account data or execute tasks without the backend. |
| Wallet identity | One-time signed login with injected EVM EOA wallets. | WalletConnect mobile transport and smart-contract wallet verification are not included. |
| Agent runtime | Hermes adapter, server-side worker, model broker, and sandbox provisioning. | Real Hermes execution, model quality, and container enforcement still need live acceptance testing. |
| Research | Controlled public-page fetching and an optional configured search service. | No interactive browser automation. Live source retrieval and search remain to be validated. |
| Images | Tool adapters for vision, generation, and editing through separately configured routes. | No live model or image request has been verified in the current delivery. |
| Memory & review | Explicit wallet/project memory; task history; pending clarification and approval controls. | Automatic learning is disabled. Approvals support Allow once or Deny; file-edit or incomplete reviews are deny-only. Live control acceptance remains to be tested. |
The source includes operator deployment guides and a validation report. Read the docs for setup and the full boundary.