Agent infrastructure

The harness the rest of the ecosystem runs on.

EvAgent is a CLI and terminal interface for running agent work: a TypeScript core that dispatches across domains, an evsmem memory server that keeps context between steps, and a setup path that gets both running in one command.

Quick start

Taken from the published package README. Commands are copied verbatim rather than paraphrased.

install

npm install -g @evsphere/evagent

If npm 12 blocks the install script, reinstall with the allow-scripts flag below. The evagent command also falls back to the platform package's binary, so it works even when install scripts were skipped.

install — if scripts are blocked

npm install -g --allow-scripts=@evsphere/evagent @evsphere/evagent

setup (one-time)

evagent -setup

Interactive: shows a banner and asks every question up front — CPU or GPU build, embedding provider, and the deriver LLM base URL, API key and model name — then installs without further prompts.

setup — unattended

evagent -setup --auto

Skips the prompts. Downloads the BGE-M3 embedding model, roughly 634 MB, so allow for that before running it.

use

evagent
evagent run "your task here"

The first line opens the interactive TUI; the second runs a single request headlessly.

What the current release does

What is documented as working today, and where the boundary sits. Anything not documented here is not claimed.

EvAgent capabilities and their documented state in version 0.7.3
Capability	Detail	State
Install	Single global npm install; per-platform binaries are resolved automatically.	Available
Setup	`evagent -setup` installs evsmem with its dependencies and embedding model, then starts it. `--auto` skips the prompts.	Available
Interactive use	`evagent` opens the terminal UI.	Available
Headless use	`evagent run "msg"` runs a single request without the TUI.	Available
Parallel dispatch	Work is dispatched across domains from a compiled core.	Available
Memory	The evsmem memory server runs locally at 127.0.0.1:9876 and holds its configuration in `evsmem/.env`.	Available
EvSystem-I	The decision layer shared with EvMem. It resolves routing and selection decisions between the harness and the memory service, and is designed to respond in roughly 20 ms.	In development
Embedding and deriver models	Setup asks for an embedding provider and a deriver LLM base URL, API key and model name. EvSphere does not supply these credentials.	Your own keys
Open-source reuse	No licence has been granted and no public repository is published. npm reports the package as UNLICENSED.	Not available

What it does

EvAgent is the platform layer of the ecosystem: a harness that dispatches work to specialised agents, gives them tools and skills, runs them from a terminal interface, and keeps memory available across steps. It is the component the other branches plug into.

Runs agent work from a terminal interface, interactively or headlessly
Dispatches work across domains from a TypeScript core.
Ships alongside an `evsmem` memory server so agents keep context between steps
Sits on top of EvSystem-I, the decision layer it shares with EvMem

Evidence

Captures from the current build. The slots below are waiting on sanitised recordings — they are not product screenshots and should not be read as any.

EvAgent terminal interface showing an architect pass breaking a broad research task into six sub-tasks, each assigned to a named specialist: technology landscape analysis, multi-agent orchestration evidence, long-term memory evidence, tool integration evidence, orchestration analysis and long-term memory analysis. — Task decomposition: a broad request split into sub-tasks and dispatched to named specialists.

EvAgent terminal interface showing the output of a research run: a landscape briefing comparing five open-source AI agent frameworks across multi-agent orchestration and long-term memory, in comparison tables with evidence anchors, followed by an extractive memory section and a stated takeaway. — Assembled output: a sourced comparison produced by the specialist passes, with evidence anchors and an explicit statement of what the evidence does not support.

How a task moves through the harness. Schematic of the intended structure — not a capture of the running system.

Source and documentation

The published package, plus a public documentation repository for this branch.

DocumentationPublic docs — no source code

Package on npmhttps://www.npmjs.com/package/@evsphere/evagent

Capability

Detail

State

Install

Single global npm install; per-platform binaries are resolved automatically.

Available

Setup

`evagent -setup` installs evsmem with its dependencies and embedding model, then starts it. `--auto` skips the prompts.

Available

Interactive use

`evagent` opens the terminal UI.

Available

Headless use

`evagent run "msg"` runs a single request without the TUI.

Available

Parallel dispatch

Work is dispatched across domains from a compiled core.

Available

Memory

The evsmem memory server runs locally at 127.0.0.1:9876 and holds its configuration in `evsmem/.env`.

Available

EvSystem-I

The decision layer shared with EvMem. It resolves routing and selection decisions between the harness and the memory service, and is designed to respond in roughly 20 ms.

In development

Embedding and deriver models

Setup asks for an embedding provider and a deriver LLM base URL, API key and model name. EvSphere does not supply these credentials.

Your own keys

Open-source reuse

No licence has been granted and no public repository is published. npm reports the package as UNLICENSED.

Not available