Comparison

Nairi vs Viktor

Two AI agents that live in Slack. Different stances on when the agent acts, what it can run, and how much you can shape it. Here’s an honest side-by-side.

Try Nairi free See Slack integration

TL;DR

Both products help teams use AI agents inside Slack. Two practical differences. First, the trigger model: Nairi only acts on an explicit mention or scheduled job; Viktor’s product is built around proactive engagement across channels. Second, the execution environment: Nairi agents run in a full isolated container with shell, file system, and build tools, so the agent can clone your repo, build your code, run tests, and use any CLI you give it. Viktor runs inside fixed product workflows.

Pick Nairi if you want an agent your team can shape: choose what it can reach, give it your build tools, layer your own skills and MCP servers on top. Pick Viktor if you want a workspace coordinator that takes initiative inside its productized lanes.

Side by side

What evaluators usually want to know before installing either product.

Dimension

Nairi

Viktor

Trigger model

Only acts when directly mentioned or triggered by a scheduled job you configured.

Designed to proactively reach out to teammates and act across channels without an explicit mention.

Execution environment

Each agent runs in its own isolated container with a full sandbox: shell, file system, package manager, build tools. The agent can clone your repo, run your build, execute tests, install dependencies, and use any CLI you give it access to.

Runs inside fixed product workflows. No general code execution sandbox published. The agent does what the product was built to do, not arbitrary tasks.

Customization

Per-agent rules, skills, MCP servers, env vars, and connected repos. Tune what each agent can do, see, and reach. You build the agent your team needs.

Productized workflows. Limited surface for changing what the agent does outside of vendor-supplied capabilities.

Channel access

You pick which channels Nairi can see at install. Changeable from workspace settings any time.

Public reports describe access to channels beyond what users granted at install.

DM behavior

Won’t DM teammates unless someone explicitly mentions Nairi in a DM or a scheduled job is configured to post one.

Active outbound DMs to teammates are part of the product’s growth loop, per public user reports.

Model choice

Claude (Sonnet 4.6, Opus 4.7), GPT-5 via Codex, or open-weights via OpenCode. Configurable per agent.

Closed model selection. Not user-configurable.

Tool integration

Model Context Protocol (MCP) servers. Bring your own MCPs for GitHub, Linear, Jira, Notion, Postgres, internal APIs, anything.

Built-in integrations only. No MCP server support published.

Secrets handling

Vault-backed. A secret proxy injects credentials at runtime when an authorized agent requests them. Chat surface never sees them.

Standard OAuth scopes; secrets stored on Viktor’s infrastructure.

Self-host option

Agent containers can run on your own servers via the open-source nairid daemon. Backend stays managed.

SaaS-only.

Scheduled jobs

First-class. Any prompt on any cron expression, output posts to a channel you choose.

Limited scheduling primitives published.

Viktor descriptions reflect public information as of June 2026. Verify the latest at getviktor.com.

The deciding question: when should the agent act?

Slack-native agents make a real product decision that most other categories don’t face: should the agent take initiative, or only respond?

Viktor picks initiative. The agent is designed to engage teammates without being asked, surface context across channels, and act as a workspace-wide coordinator. Teams who want an agent that nudges and coordinates tend to like this.

Nairi picks responsiveness. Agents act only when explicitly mentioned or when a scheduled job you defined fires. Channels are pre-selected at install. DMs aren’t initiated by the bot. This is a deliberate constraint, not a missing feature.

We pick the constraint because the people deciding whether to deploy an AI agent in production Slack workspaces are usually IT, security, or the engineering lead, and they answer to compliance. Least-privilege access and explicit-trigger semantics are the easier sell in that room. When an evaluator looks at a public report about an agent accessing channels users didn’t grant, the conversation often ends right there.

That doesn’t make Viktor’s choice wrong. It makes it different. If your team values proactive coordination over explicit control, Viktor is the closer fit. If you want a tool engineers reach for and step away from cleanly, Nairi is.

What the agent can actually do

The other large difference between the two products is what runs underneath the chat surface.

Each Nairi agent runs in its own isolated container with a full sandbox. The agent has a shell, a file system, a package manager, and the ability to install any CLI you point it at. That means an engineer can ask Nairi to clone the repo, install dependencies, run the build, execute the test suite, run your linter, hit your internal CLI, generate a report from a SQL query, fix a flaky test, or stand up a one-off tool for a specific workflow.

Viktor runs inside fixed product workflows. The agent can do what the product was built to do, and that's the surface area. If your team needs a behavior the product doesn't ship, you wait for the vendor or you work around it.

This matters for two reasons. First, it's how engineering teams actually use AI agents on real codebases: they don't want a chat interface, they want something that can build, test, and ship. Second, it's what makes the agent customizable. You pick the rules, the skills, the MCP servers, the connected repos, the env vars. Every team configures their agent differently, because every team's codebase and workflow is different.

The downside of the sandbox is real: an agent that can run arbitrary code is more powerful than one that can't, which is exactly why the trigger-and-permission model from the previous section matters so much. Nairi pairs the broad capability with tight controls on when and how it's invoked.

What Viktor does well

Coordination polish. Threading follow-ups, looping in the right person, surfacing decay across channels. If you want a workspace-wide coordinator more than a code agent, that’s the focus.
Strong funding signal. A $75M Series A in May 2026 means a long runway for product investment.
Onboarding speed. Setup is fast, like Nairi. Both products treat install friction as a first-class concern.

Common questions

What teams ask before picking between the two.

Viktor (getviktor.com) is a Slack-native AI agent that announced a $75M Series A in May 2026. It positions itself as an AI coworker that proactively engages with teammates and helps coordinate work across the workspace.

Nairi only acts on an explicit mention in a channel or thread, or a scheduled job you set up in advance. Viktor’s product is designed around proactive engagement, meaning the agent reaches out to teammates and crawls channels without being asked. Both approaches are valid product choices; they appeal to different buyers. IT, security, and compliance teams typically prefer the explicit-trigger model.

Yes. Each Nairi agent runs in its own isolated container with a shell, file system, and package manager. The agent can clone your repo, install dependencies, run your build, execute tests, run linters, hit any CLI you give it, generate reports from SQL queries, and stand up one-off tooling for specific workflows. This is the main thing that separates Nairi from chat-only agents: you can put real engineering work on it, not just conversation. Viktor doesn’t publish a comparable execution environment.

Per agent you can configure: which channels it sees, which repos it can read or write to, which MCP servers it can call, which skills it can run, which env vars it has, which secrets in your vault it can request, and a free-form rules document that sets its behavior. Different agents in the same workspace can have completely different configurations. One agent for on-call ops with read-only Postgres MCP, another for shipping code with full repo access and your build pipeline.

With Nairi you pick which channels the bot can see during install, and you can revoke or change that any time. Public user reports about Viktor describe behavior in channels users didn’t grant access to (see Ben Vinegar’s May 2026 thread for the most-cited example). If your evaluators care about least-privilege access, the explicit-channel model is the lower-friction sell.

Yes. Nairi’s Slack install takes around 30 seconds. There’s no migration step: you install the app, pick channels, connect tools (GitHub, Linear, Jira, Notion, or any MCP server), and start mentioning the bot. Running both side by side for a week before deciding is the most common pattern we see.

Viktor’s focus is workspace-wide coordination: looping people in, nudging follow-ups, surfacing context across channels. Teams that want an agent that takes initiative tend to lean toward this model. Nairi takes the opposite stance by design; we treat the agent as a tool that engineers explicitly invoke, not a coworker that initiates work.

Nairi plans start at $20/month. Slack is included on every tier. See the pricing page for current tiers and limits. Viktor’s pricing isn’t published publicly at time of writing; talk to their sales team for a quote.

Try Nairi in Slack

Install the Slack app, pick your channels, mention @Nairi. Two minutes, no infra to deploy.

Try Nairi free Slack integration overview

Slack integration Pricing Documentation