nima docs

From customer signal to shipped pull request

nima is headless. Raw feedback becomes ranked evidence, a codebase-grounded spec and a tracked pull request, and you work that pipeline in Slack, from coding agents over MCP (https://mcp.getnima.io/mcp) and over the REST API (https://api.getnima.io). The web app is an admin panel for sign-in, integrations and settings.

# 1. add the marketplace (inside Claude Code)
/plugin marketplace add getnima/skills

# 2. install the plugin (in your shell), then run /mcp and sign in
claude plugin install nima@nima --scope project

Quickstart Use nima in Slack Connect an agent API reference

the loop

One pipeline, five stages

get signal in

Choose how signals reach nima

Start with whichever your feedback already lives in. All three land in the same pipeline.

no code

Connect an integration

Slack, GitHub, Linear, ClickUp, Granola, PostHog and Sentry become signals once connected in the admin panel. HubSpot is rolling out: ask the nima team. Jira connects for delivery and for ticket replies you approve, not as a signal source. Notion can be connected, but nima does not read from it or write to it yet.

any tool

POST JSON to a webhook

An owner or admin creates an endpoint from a signed-in session (API keys cannot), or asks the nima team. Then any tool can send it a JSON object: a support desk, a cron job, an alerting channel.

agents and scripts

File signals with the API or MCP

POST /v1/signals with a key, or the create_signal tool from Claude Code and Codex. Same route, same scope, same workspace boundary.

work with nima

Where people and agents work the pipeline

Nothing in the pipeline needs a web page. Each surface below is a way to read or act on the same records.

people

The Slack assistant

Mention nima or message it directly to ask about signals, beacons and specs. When the work ships, nima replies in the thread the report came from.

agents

MCP tools for coding agents

Tools grouped by area: knowledge, pipeline, watches, operations and outbound actions. One OAuth consent screen, scoped to your workspace.

workspace setting

Watches and reminders

Once an owner or admin turns watches on, get a Slack message when a beacon is ready, a spec needs a decision or a PR merges. Scheduling a message to yourself works in every workspace with Slack linked.

workspace setting

Outbound actions with approvals

Once an owner or admin turns outbound actions on for the workspace, nima drafts a ticket reply or a question to a teammate, and a person approves it on a Slack card before it is sent. Reminder DMs to members go out directly, and owners and admins decide which actions need approval.

agents

Agent operations

Hand a task to a coding agent and follow it as an operation, from dispatch to a reviewable result.

admins

What is on in your workspace

Which features are on for every workspace, which an owner or admin turns on, and which the nima team enables on request.

get knowledge out

Hand nima to your coding agent

The knowledge graph holds what a codebase cannot: who asked, what they said, what was decided and rejected. Your agent can ask it before it writes a line.

why and who

ask_knowledge_graph

Who asked for a change, what they said and what was decided.

where

search_code

Ranked files with summaries, symbols and line ranges, plus the commit the index is at.

impact

blast_radius

What a change to a file or symbol can reach, before the agent edits it.

schema

get_data_model

The tables and fields your code defines, parsed from ORM models and SQL migrations.

The skill teaches the agent when to ask the graph, how to phrase the question and how to answer without overclaiming. Claude Code gets it through the nima plugin; Codex and other agents install it with npx skills add getnima/skills. Setup for each agent is in Connect an agent.

example

A signal, end to end

What one piece of feedback looks like as it moves through the loop.

A teammate posts a customer's report in a Slack channel nima is in: bulk export fails on their workspace. nima classifies the message as a bug, extracts the intent and tags the origin internal, because Slack messages come from your own team.

{
  "source": "slack",
  "classification": "bug",
  "origin": "internal",
  "extractedIntent": "Large CSV exports time out before completing",
  "status": "ready"
}