Nairinairi

AI agents in Slack, without the context switching

Mention @Nairi in any channel or thread. An agent picks up the task, runs against your codebase or runbooks, and replies in-thread. Built on Claude Code, Codex, or OpenCode.

Connect Slack Read the docs

Included on every plan · no per-seat fees · your secrets stay in your vault

eng-helpSlack
D
Dana9:14 AM
@Nairi the CSV export drops rows with unicode names - LIN-482 has repro steps
Nairiapp9:15 AM
On it. Cloning acme/exports, reading LIN-482…
Nairiapp9:22 AM
Root cause: latin-1 encode in export_csv.py:88. Fixed it, added a regression test - 47/47 passing.
PRacme/exports#291 checks passing
M
Marco9:31 AM
nice - also gate it behind the exports_v2 flag
Nairiapp9:32 AM

A few things teams use Nairi for in Slack

These are starting points, not the limit. The agent runs whatever you mention it on, with whatever tools you've connected.

Ship code from a thread

Drop a Linear ticket or bug description in a thread. Nairi checks out the repo, makes the changes, runs the tests, and posts a PR link in the same thread.

Run on-call playbooks

Page Nairi from your on-call channel. It pulls logs, checks dashboards via MCP tools you configure, and correlates with recent deploys.

Triage support

Summarize the last 20 messages from a customer-success channel, draft replies, or escalate the ones that need a human.

Schedule recurring jobs

Have Nairi post a weekly PR digest or a daily incident summary at a time you pick. The output lives in your team channel, not in someone's personal Notion.

Loop the agent into existing threads

Mention @Nairi mid-conversation and it reads the full thread before responding. No re-explaining context.

+ whatever else your team does

Hand it your build script, your migrations runner, your custom CLI, the SQL helper you wrote last quarter. The agent uses what you give it.

Your Slack agent, wired to your stack

Conversations from Slack route to an agent in your org. Configure what it can reach - repos, MCP servers, skills, secrets, custom rules. Add or remove any of these any time.

Slack
@-mention in any channel, thread, or DM
nairiruntimeN AGENTS
7F3Afix-checkout-test›claude-code
2B19triage-zendeskcodex
A4E2release-notesopencode
each in its own container, all inside your org
GitHub repos
Read code, open PRs, run CI
MCP servers
1,000+ - Postgres, Linear, Notion
Skills
Reusable prompts & playbooks
Vaults
Secrets injected at runtime
Rules
Custom instructions per agent

Common workflows

A few of the patterns teams reach for on day one. The agent does much more once you wire it into your stack.

PR review and code shipping

Engineers mention Nairi with a ticket link or bug description. It checks out the repo, makes the changes, runs the tests, and opens a PR with a summary in the thread. Reviewers respond inline, and Nairi keeps full thread context.

On-call incident response

Page Nairi from your on-call channel during an incident. It pulls logs, checks dashboards via MCP tools you configure, correlates with recent deploys, and posts findings in the thread - ready for the postmortem.

Async product and eng updates

Schedule Nairi to post a recurring update: last week's shipped PRs, open incidents, customer escalations. The update lives in your team channel, not in a doc nobody reads.

How it works

Install in 30 seconds, then choose what to connect. Your secrets stay in your vault, injected at runtime via a secret proxy. The chat surface never sees them in plaintext.

  1. Install the Nairi Slack app
    One-click OAuth, around 30 seconds. You pick which channels Nairi can see during install.
  2. Connect what the agent should reach
    Pick from GitHub repos, MCP servers (1,000+ available, including Postgres, Linear, Notion, or your internal APIs), reusable skills, custom rules, and vault-backed secrets. Add or remove any of these any time.
  3. Mention @Nairi in any channel or DM
    Nairi spins up an isolated agent container, runs the task with Claude Code, Codex, or OpenCode (you choose), and replies in-thread.
  4. The conversation persists
    Follow up, redirect, or ask for changes. The agent remembers the full thread context.

Get started in two minutes

Sign up, click "Add to Slack", pick your channels, and mention @Nairi. No infra to deploy. No keys to rotate. No agents to babysit.

Try Nairi free View pricing

Questions about Slack

The things teams actually ask before installing.

No. During install you pick exactly which channels Nairi can see. You can change this any time from the workspace settings or by reinstalling the Slack app.
No. Nairi only acts when it's directly mentioned in a channel or thread, or triggered by a scheduled job you configured. It doesn't crawl channels, doesn't initiate DMs, and doesn't take action without an explicit trigger.
Only what you grant. During install you pick which Slack channels Nairi can join. In those channels the agent reads messages it’s mentioned in plus the surrounding thread context, so it can follow up properly. It doesn’t crawl other channels, doesn’t scan DMs it wasn’t pulled into, and doesn’t train on your message history. Full scope at nairi.ai/docs/help/security.
No. Nairi can’t add itself to a channel it wasn’t invited to, can’t expand its own permissions, and can’t spawn new agents in other workspaces on its own. Every channel grant requires a human with the right Slack permissions. The Slack-agent trust thread from May 2026 surfaced cases where a chat-native bot acted in channels it wasn’t invited to. Nairi’s permission model doesn’t allow that kind of behavior.
Same way Slack does. Remove the bot from a channel and the agent loses access on the next event. Uninstall the Nairi Slack app and Slack revokes the workspace token immediately. Vault secrets you connected stay in your vault. Slack revocation doesn’t touch them, and you can rotate or delete them independently. Lifecycle details at nairi.ai/docs/help/security.
Yes. Every conversation is recorded with timestamp, prompt, model used, tools called, secrets requested, files touched, and token cost. You can browse it in the dashboard or pull it via the public API. The message schema lives at nairi.ai/docs/api/conversations/message-reference. Broader security model: nairi.ai/docs/help/security.
Both. Nairi keeps full thread context, so you can mention it mid-thread and it picks up where the conversation left off. Replies post in-thread by default.
You choose: Claude via Claude Code, GPT via Codex, or open-weights via OpenCode. Configuration is per agent or per channel.
Secrets live in your Nairi vault, never in chat. At runtime a secret proxy intercepts outbound HTTPS calls from the agent and injects credentials only when an authorized agent requests them. Slack never sees them.
Yes. Agent containers can run on your own servers via the open-source nairid daemon. The backend remains hosted by Nairi, so you don't manage the orchestration layer.
Yes. The Scheduled Jobs feature runs prompts on a cron expression. The result posts to a channel you specify. Common uses: daily digests, weekly reports, periodic health checks.
Every assistant message exposes cost_usd, input_tokens, output_tokens, cache_read_tokens, cache_write_tokens, and model. Pull the messages from the public API and aggregate per agent. The buyer-facing dashboard view is shipping next. Honest caveat: the Claude Code and OpenCode harnesses report these fields; the Cursor harness doesn’t yet. Field reference: nairi.ai/docs/api/conversations/message-reference.
Slack is included on every plan. See the pricing page for current tiers.

Related

On-call incident responseAutomated PR reviewCompany brain on SlackData analyst on SlackDiscord integrationGitHub integrationNairi vs Cursor Cloud AgentsNairi vs Claude Code for SlackWhy every AI coding agent is built for one userPricingDocumentation