# Nairi

> Team-shared AI agent platform. Configure agents once and deploy them via Slack, Discord, web, or REST API.

Nairi documentation is hosted at https://nairi.ai/docs/help. A grouped index of every docs page is at https://nairi.ai/docs/llms.txt and a combined plain-text dump of every page is at https://nairi.ai/llms-full.txt. Append `.md` to any docs page URL to get the raw Markdown variant (e.g. https://nairi.ai/docs/help/getting-started.md).

## Site

- [Homepage](https://nairi.ai): product overview and signup
- [Documentation](https://nairi.ai/docs/help): help center and REST API reference
- [Blog](https://nairi.ai/blog): essays from the Nairi team on multi-vendor AI agents, context portability, and the platform layer above LLMs
- [Changelog](https://nairi.ai/changelog): release notes
- [Slack integration](https://nairi.ai/integrations/slack): how Nairi runs as an AI agent inside Slack
- [GitHub integration](https://nairi.ai/integrations/github): how Nairi runs as an AI agent across GitHub repos, PRs, and Issues
- [Discord integration](https://nairi.ai/integrations/discord): how Nairi runs as an AI agent inside Discord servers
- [Nairi vs Cursor Cloud Agents](https://nairi.ai/compare/nairi-vs-cursor-agents): side-by-side comparison covering Discord support, open-source self-host, multi-harness choice, and where Cursor wins
- [Nairi vs OpenClaw](https://nairi.ai/compare/nairi-vs-openclaw): side-by-side comparison framing Nairi as "OpenClaw for teams" with org-scoped vaults, RBAC, shared agents in Slack/Discord channels, and where OpenClaw wins (open-source, 20+ chat apps, personal voice)
- [Nairi vs Claude Tag](https://nairi.ai/compare/nairi-vs-claude-tag): side-by-side comparison vs Anthropic's Claude Tag Slack teammate (beta, June 23, 2026, Claude Team/Enterprise only, Opus 4.8) covering context lock-in vs workspace-scoped portable context, multi-vendor harness and model choice (Claude Code/Codex/OpenCode) vs Anthropic-only, Slack + Discord vs Slack-only, open-source self-hostable runtime vs Anthropic-cloud-only, and where Claude Tag wins (ambient mode, frictionless install for existing Claude teams, first-party Opus 4.8, bundled pricing)
- [Nairi vs Claude Code for Slack](https://nairi.ai/compare/nairi-vs-claude-code-for-slack): side-by-side comparison vs Anthropic's official Slack integration covering per-org vs per-user agents, harness choice (Claude/Codex/OpenCode), Discord support, org vault for secrets, self-hostable runtime, and where Claude Code for Slack wins (frictionless install, cloud VM toolchain, auto-fix PR loop, session handoff)
- [Nairi vs Claude Managed Agents](https://nairi.ai/compare/nairi-vs-claude-managed-agents): side-by-side comparison vs Anthropic's hosted agent runtime covering zero-code Slack/Discord self-serve vs build-your-own-bot, multi-harness choice vs Claude-only, full self-hosted agent loop vs Anthropic-orchestrated, and where Claude Managed Agents wins (hosted infra breadth, AWS Marketplace, polyglot SDKs, consumption pricing)
- [Nairi vs Devin](https://nairi.ai/compare/nairi-vs-devin): side-by-side comparison vs Cognition's Devin covering self-host the full agent loop vs a hosted-cloud agent, harness choice in shared-chat agents (Claude Code/Codex/OpenCode vs Devin's own harness), pricing predictability (LLM pass-through vs ACU quotas), Discord support, and where Devin wins (Devin Review PR specialist, Auto-Triage turnkey channel monitor, hosted infra breadth, curated MCP marketplace, Devin Desktop polish, SOC2 Type II enterprise procurement)
- [Nairi vs Lindy](https://nairi.ai/compare/nairi-vs-lindy): side-by-side comparison vs Lindy's personal AI work assistant (inbox/calendar/scheduling, per-inbox pricing from $49.99/mo + monthly credit pool, 2,500+ pre-built integrations) covering shared team-channel agent vs single-person assistant, MCP-first tools vs curated catalog with no MCP support, harness and model choice (Claude Code/Codex/OpenCode) vs proprietary hosted runtime, open-source self-hostable nairid runtime vs cloud-only closed source, flat per-org pricing vs per-inbox credits, and where Lindy wins (faster start for non-technical users, mature email/calendar workflows, ~170 reviews at 4.9/5 on G2)
- [Nairi vs Zapier AI](https://nairi.ai/compare/nairi-vs-zapier-ai): side-by-side comparison vs Zapier's workflow-automation platform and Zapier Agents (formerly Zapier Central), covering agent-first runtime vs automation-first engine with an AI layer, shared @mentionable Slack/Discord channel agent vs Zapier's web-chat/Chrome-extension agent where Slack is only a send destination, MCP-first tools vs a 9,000+ pre-built app catalog, harness and model choice (Claude Code/Codex/OpenCode) vs Zapier's hosted runtime (BYO-model via AWS Bedrock on Enterprise), open-source self-hostable nairid runtime vs cloud-only closed source, flat per-org pricing from $20/mo vs per-task billing (each action step + connector call = 1 task, MCP tool call = 2 tasks; plans from Free to Professional $19.99/mo to Team $69/mo), and where Zapier wins (unmatched 9,000+ connector catalog, easiest visual builder for non-engineers, deterministic Zaps when exactness matters)
- [Use cases index](https://nairi.ai/use-cases): the four documented use-case playbooks Nairi ships (on-call incident response, automated PR review, company brain on Slack, data analyst on Slack), each with a demo video
- [On-call incident response use case](https://nairi.ai/use-cases/oncall-incident-response): how Nairi runs as an on-call agent in the incident channel, pulling metrics, tailing logs, correlating against recent deploys, drafting status updates and postmortems, with read-only-by-default safety and org-vault credentials
- [Automated PR review use case](https://nairi.ai/use-cases/automated-pr-review): how Nairi runs as a first-pass PR reviewer via a GitHub Action calling the REST API, with per-repo skills and rules, harness choice (Claude Code/Codex/OpenCode), and a self-hostable runtime so code never has to leave your infra
- [Company brain on Slack use case](https://nairi.ai/use-cases/company-brain): how Nairi runs as a shared agent connected to Notion/Confluence/Drive answering "how do we do X here?" with citations, plus action MCPs (Linear/Jira/GitHub/Slack writes), confidentiality rules, and a docs-maintenance loop
- [Data analyst on Slack use case](https://nairi.ai/use-cases/data-analyst): how Nairi runs as a plain-English data analyst on Postgres/Snowflake/BigQuery via MCP, with schema-cheat-sheet rules, three layers of safety (read-only DB creds + restricted MCP + query-safety rules), and scheduled-job KPI digests

## Documentation

- [Full docs index](https://nairi.ai/docs/llms.txt): grouped list of every help and API page
- [Full docs dump](https://nairi.ai/llms-full.txt): every page concatenated as plain text

## Blog

- [Why you should be able to self-host your agent runtime (2026-08-01)](https://nairi.ai/blog/2026-08-01-self-host-agent-runtime): In two months, two vendors walked back self-hosting and harness freedom. Cognition put self-hosted Devin into maintenance mode. Anthropic blocked third-party harnesses on the Claude Code subscription. If your agent runtime depends on one vendor's billing model, regional availability, or ban policy, your team is one announcement away from a forced migration. Here is what a self-hostable runtime actually has to give you, and where Nairi is honest about the line between what runs on your hardware and what does not.
- [Why every AI coding agent is built for one user (2026-07-25)](https://nairi.ai/blog/2026-07-25-single-user-ai-agents): Claude Code, Cursor agents, OpenCode, Devin. Every AI coding agent on the market is built for one developer at a time. That is an architectural choice, not a missing feature, and it holds until the moment a second human needs to read or continue what the agent did. Here is where the single-user assumption breaks and what multi-user actually requires.
- [The trust ceiling of Slack-native AI agents (2026-07-18)](https://nairi.ai/blog/2026-07-18-trust-ceiling-slack-ai-agents): In 2026, AI agents in shared Slack workspaces hit their first wave of public trust incidents: a worm-like growth hack, org-wide bans, a supply-chain backdoor, a privilege-escalation CVE. None of them are one vendor being reckless. They are the same missing primitive, a permissions model a buyer can actually reason about. Here is the pattern, what buyers are asking for, and what Nairi ships today.
- [What predictable agent pricing looks like, and why nobody is shipping it (2026-06-27)](https://nairi.ai/blog/2026-06-27-predictable-agent-pricing): Devin's ACUs, Cursor's compute budget, Lindy's per-seat fees, Zapier's tasks, Viktor's credits, Claude Code's weekly limits. Every flagship AI agent has at least one viral pricing complaint. Here's what predictable agent pricing looks like instead, and where Nairi lands.
- [Claude Tag and the context lock-in problem (2026-06-24)](https://nairi.ai/blog/2026-06-24-claude-tag-context-lock-in): Anthropic shipped Claude Tag this week. The product is good. The architecture it pushes you toward is the part worth thinking about before you adopt.

## Changelog

- [Scoped & Expiring API Keys (2026-09-15)](https://nairi.ai/changelog/2026-09-15-scoped-api-keys): API keys now carry scopes and an optional expiry, so a key for reading conversations can't quietly delete your agents.
- [Activity History (2026-09-08)](https://nairi.ai/changelog/2026-09-08-activity-history): See who changed what, and when. Every configuration change in your org is now recorded and visible, with the actor, the field, and the before/after value.
- [Roles & Permissions (2026-09-01)](https://nairi.ai/changelog/2026-09-01-roles-and-permissions): Members now come in two roles — admin and member — so you can invite people without handing over the keys to everything.
- [Multiple Organizations (2026-08-25)](https://nairi.ai/changelog/2026-08-25-multiple-organizations): One account, many organizations. Create separate orgs, switch between them from the sidebar, and accept invites without leaving the app.
- [Invite Your Team (2026-08-18)](https://nairi.ai/changelog/2026-08-18-invite-your-team): Nairi is no longer a single-seat product. Invite colleagues into your organization, see everyone on a members page, and manage who's in.
- [Public API: Hardening Pass (2026-06-10)](https://nairi.ai/changelog/2026-06-10-public-api-hardening): A round of validation, error response, and consistency fixes for the public API. Three new validations may reject requests that used to silently succeed.
- [Public API: Per-Message Cost and Tokens (2026-06-03)](https://nairi.ai/changelog/2026-06-03-per-message-cost-and-tokens): Assistant messages on the public API now carry cost_usd, input/output/cache token counts, and the model used so you can track spend per turn.
- [Public API: File Attachments (2026-05-27)](https://nairi.ai/changelog/2026-05-27-public-api-attachments): Send files to a Nairi agent and read back any files it produces through the public API, plus an `attachment_ids` field on conversation start and continue.
- [Help Center and API Docs (2026-05-21)](https://nairi.ai/changelog/2026-05-21-help-center-and-api-docs): A new help center with guides for operators, builders, and team leads, plus full API reference docs for everyone driving Nairi from code.
- [Public API: Incremental Conversation Sync (2026-05-12)](https://nairi.ai/changelog/2026-05-12-conversation-sync-api): The list-conversations endpoint now supports time-ordered pagination, updated_after filtering, and returns all job statuses.
- [Task History (2026-05-05)](https://nairi.ai/changelog/2026-05-05-task-history): Completed tasks now stick around. Switch between Active and Completed tabs to see what your agents finished, what failed, and what was abandoned.
- [Connected Channels on Agent Cards (2026-04-28)](https://nairi.ai/changelog/2026-04-28-connected-channels): See which Slack and Discord channels each Nairi agent is connected to, right on the dashboard. Plus Slack forwarded messages, share link cleanup, and more.
- [MCP Tool Reliability Fixes (2026-04-21)](https://nairi.ai/changelog/2026-04-21-mcp-reliability-fixes): Several fixes for MCP tool connections from the marketplace. Re-authentication, cleanup, and compatibility improvements across the board.
- [Multiple API Keys (2026-04-14)](https://nairi.ai/changelog/2026-04-14-multiple-api-keys): Create multiple named API keys per organization. Separate keys across environments, team members, or integrations and revoke them independently.
- [Multi-User Awareness (2026-04-07)](https://nairi.ai/changelog/2026-04-07-multi-user-awareness): Agents now know who's talking to them. When multiple team members interact with the same agent, it sees each person's name and can give personalized responses.
- [Custom System Prompts (2026-04-02)](https://nairi.ai/changelog/2026-04-02-custom-system-prompts): Customize the system prompt for each agent. Control exactly how your agent behaves, what tone it uses, and what instructions it follows.
- [Reliable Message Delivery (2026-03-31)](https://nairi.ai/changelog/2026-03-31-reliable-message-delivery): Messages between you and your agent are now delivered through a redundant channel. Brief network interruptions no longer cause lost messages.
- [File Attachments (2026-03-26)](https://nairi.ai/changelog/2026-03-26-file-attachments): Agents can now send files back to you. Generated reports, CSVs, images, or any file - delivered directly to Slack, Discord, or the web dashboard.
- [Live Agent Progress (2026-03-24)](https://nairi.ai/changelog/2026-03-24-agent-progress): See what your agent is doing in real time. Tool calls, reasoning steps, and subtasks stream to the UI as they happen.
- [App Scheduled Jobs (2026-03-19)](https://nairi.ai/changelog/2026-03-19-app-scheduled-jobs): Scheduled jobs no longer need a channel. Create app-level jobs that run on a cron schedule and pick which agent handles them.
- [Skill Marketplace (2026-03-17)](https://nairi.ai/changelog/2026-03-17-skill-marketplace): Search and install pre-built skills from a growing catalog. One click to add a skill to your agent, or download it for local use.
- [Rich Slack Formatting (2026-03-12)](https://nairi.ai/changelog/2026-03-12-rich-slack-formatting): Agent responses in Slack now render with native tables, proper emoji, and better handling of long messages. No more walls of monospace text.
- [Meet Nairi (2026-03-10)](https://nairi.ai/changelog/2026-03-10-nairi-rebrand): eksec is now nairi. New name, new look, same platform. Your agents, integrations, and workflows continue to work exactly as before.
- [Tool Marketplace (2026-03-07)](https://nairi.ai/changelog/2026-03-07-tool-marketplace): Browse a catalog of MCP tools and install them to your agent with one click. OAuth and API key auth flows are handled for you.
- [Redesigned Dashboard (2026-03-05)](https://nairi.ai/changelog/2026-03-05-new-dashboard-design): The nairi dashboard got a fresh look. Easier to find what you need, less clutter, and a smoother experience across the board.
- [Custom Environment Variables (2026-03-03)](https://nairi.ai/changelog/2026-03-03-environment-variables): Set custom environment variables on your agents directly from the dashboard. They're applied on startup and available to all processes inside the container.
- [OpenAI Codex Support (2026-02-26)](https://nairi.ai/changelog/2026-02-26-codex-integration): nairi agents can now run on OpenAI's Codex CLI. Set up Codex as your agent's coding harness and everything else stays the same.
- [MCP Tool Proxy (2026-02-24)](https://nairi.ai/changelog/2026-02-24-mcp-tool-proxy): Your agent's MCP tool connections now run through a proxy that hides your secrets from the agent. It can use your tools without seeing the credentials.
- [nairi.ai is live (2026-02-10)](https://nairi.ai/changelog/2026-02-10-nairi-ai-is-live): Build an AI agent once and share it with your team. Connect Claude Code or OpenCode, add your tools and skills, then deploy via Slack, Discord, or web.

## Optional

- [Privacy policy](https://nairi.ai/privacy)
- [Terms of service](https://nairi.ai/terms)
- [Data processing agreement](https://nairi.ai/dpa)