---
title: What can Nori do?
description: Nori can run coding agents from team surfaces, connect them to repositories and tools, automate recurring work, and preserve sessions across machines.
status: verified
sidebar:
  order: 2
---

Nori can run a supported coding agent in an isolated cloud session, give it
controlled access to the context it needs, and return the work to the teammate or
system that requested it.

## Start agent work from the tools your team already uses

Managed Sessions accept work from several surfaces:

| Surface | Typical use |
|---|---|
| Slack | Mention the bot in a channel, continue in its thread, attach files, and use commands to inspect or control the session |
| Web dashboard | Start a chat, review session history, resume work, and administer the fleet |
| Discord | Start and continue agent conversations in supported channels |
| Handroll | Work interactively in a terminal and hand a session between local and cloud contexts |
| Schedule or reminder | Run maintenance, reports, and follow-ups without a person online |
| Webhook | React to an external event with a new or continuing agent session |

All surfaces reach the same broker and fleet. A conversation can be reviewed or
resumed from another surface when the session supports it.

## Give agents a real working environment

A session machine can receive:

- One or more workspace repositories.
- An organization setup script for dependencies and environment preparation.
- A selected coding agent, model, and skillset.
- GitHub access through an installed GitHub App or a personal access token.
- Google Workspace access for the signed-in user.
- Managed MCP connections and custom environment variables.
- Optional infrastructure credentials when an administrator enables them.

This makes Nori suitable for repository work, operational investigation,
documentation, analysis, and other tasks that depend on tools beyond a chat box.

## Delegate software work

The exact result depends on the selected agent and granted integrations, but
common tasks include:

- Diagnose and fix a failing test or CI job.
- Implement a scoped feature and prepare a pull request.
- Update dependencies and verify the build.
- Investigate an error using source code and observability data.
- Review a repository and explain an unfamiliar subsystem.
- Refresh documentation after a product change.
- Run a recurring repository health check.

Agents still follow approval and access boundaries. Giving an agent a task does
not silently grant credentials or permissions that the organization has not
configured.

## Preserve and resume work

Nori records transcripts and checkpoints so the logical session can outlive one
agent process or machine. In Slack, a reply can resume an inactive session unless
the thread was explicitly released with `!done`. In the dashboard, resumable
sessions are listed with their history.

See [checkpoints and resume](/sessions/concepts/checkpoints-and-resume/) for the
durability model and [session lifecycle](/sessions/reference/lifecycle/) for the
exact state transitions.

## Standardize agent behavior

A [skillset](/skillsets/) packages the instructions, skills, and configuration a
team wants an agent to use. Managed Sessions can apply a default skillset at
provision time, individual sessions can override it, and triggers can select one
for unattended work.

## Run on managed or deployment-specific compute

Hosted Nori fleets use Modal for new fleet deployments, while Sprites remains
supported for existing fleets. Dedicated Nori deployments can also use AWS EC2
for bring-your-own-cloud infrastructure or local Docker workers. Those are
deployment choices, not per-session toggles.

Review the [support matrix](/sessions/reference/supported-agents-and-compute/) for
current availability and [fleets and compute](/sessions/concepts/fleets-and-providers/)
for the architecture.

## Use the open-source tools locally

If you do not need managed cloud sessions, the [Nori CLI](/cli/) provides a local
ACP host for Claude Code, Codex, Gemini CLI, and custom agents. [Nori
Skillsets](/skillsets/) manages portable agent configuration across supported
tools. Both projects are public on GitHub.

## Current boundaries

- Agent support differs between managed Sessions and the open-source CLI.
- Compute backends are selected for a fleet or deployment, not for each prompt.
- Integration access depends on organization configuration and user permissions.
- A session can automate work, but consequential external actions remain subject
  to the agent's approval policy and the permissions available to it.

For product boundaries and common evaluation questions, continue to the
[FAQ](/resources/faq/).
