Deploy agents into real operations.

vespid.ai is a product and infrastructure site about one problem: how agents get deployed into real operator workflows without losing control, auditability, or runtime safety.

Read the documentation

• Product surfaces
• Runtime control

1 live product lanevespid_voice is the current user-facing system for realtime orchestration, approvals, and backend routing.

5 active layersProduct surface, delegated access, runtime substrate, profile portability, and search/discovery tooling are being shaped as one stack.

3 public entry pointsProjects, documentation, and blog posts give the site a familiar reading path instead of a custom taxonomy.

Featured project

vespid_voice

An Android-first voice product where orchestration, confirmations, realtime execution, and agent routing have to work in a real user loop.

voice interfaceapproval policyruntime routing

Discovery

Open geo-skill

Go straight to the GEO / AI-search layer if you care about how the public stack gets found, cited, and distributed across machine-readable search surfaces.

Documentation

Read the control model

Use the docs to understand approvals, runtime boundaries, operator state, and how the pieces fit together.

Discovery surface

Make the public stack legible to AI-native search

geo-skill is the part of the stack that turns public discoverability into operational work: AI search positioning, llms.txt coverage, structured data, and machine-readable site/repo surfaces that can be audited instead of guessed.

Featured discovery layer

geo-skill

A GEO skill pack and Python CLI for AI search, ChatGPT Search, llms.txt, structured data, and distribution surfaces that help the public stack get found and cited.

AI searchllms.txtstructured data

Release

Open the current tagged release

Jump straight to geo-skill v0.4.0 if you want the latest public batch and release notes.

Project page

See how discovery fits the stack

Read the full project page if you want the GEO / AI-search layer in the same systems framing as runtime, authorization, and portability.

Platform overview

What is being built

The site is organized like a typical technical platform: one featured product, a small set of core systems, and a documentation path that explains how they work together.

Product

Live lane

vespid_voice

Voice-first operator interface for realtime orchestration, confirmations, and agent bridges.

androidvoice UXagent bridge

The clearest entry point if you want to see the stack at the product layer.Open project

Authorization

Reference v0

SkillAuth

Delegated access for agents working against login-gated systems without inheriting raw human credentials.

scoped accessaudit trailrevocable capability

Public repoOpen project

Runtime

Public core

vespid

Execution primitives for local sandboxing, brokered control, session storage, and vault-backed state.

sandbox runtimebroker controlstate handling

Public repoOpen project

Portability

Incubating

hermes-profile-sync

Export, sanitize, sync, and restore Hermes profiles without treating ~/.hermes as one unsafe blob.

profile exportmanifest diffsafer restore

Operator portability as part of the trust model.Open project

Discovery

Released CLI

geo-skill

GEO skill pack and Python CLI for AI search, ChatGPT Search, llms.txt, structured data, and machine-readable public discovery.

AI searchllms.txtdistribution

Public repoOpen project

Start in the right place

Projects

Browse the stack

Start here if you want the shortest path to the actual product and infrastructure pieces.

Documentation

Understand the control layers

Read the durable reference material for runtime control, delegated access, approval policy, and system structure.

Blog

Follow technical decisions

Launch notes, architecture writeups, and deployment lessons that have not yet hardened into documentation.

Why this structure

01Lead with the product

The featured project shows where the ideas actually meet users instead of staying abstract.

02Explain the platform

Projects and docs separate the running systems from the reusable operating model behind them.

03Keep updates readable

The blog carries in-progress decisions and launch context without turning the whole site into a journal.