# Runbear

Aiolum story: https://aiolum.com/story/runbear/ · added 2026-10-23 · updated 2026-10-23

**In short.** Runbear (AI agents inside Slack and Microsoft Teams, USA): US$400K annual recurring revenue reported, October 2025. Evidence grade D: reported, not verified. Team: 2 co-founders; 5 people listed by a directory in 2026; headcount at the figure not stated in any source read. Launched 23 November 2023; still running, revenue growing.

AI agents inside Slack and Microsoft Teams · USA · AI-native · team of 2 to 10 · still running, revenue growing · group: Tools for using AI models

- Outcome: US$400K annual recurring revenue reported, October 2025 (evidence grade of the headline number: D; headline in US dollars: 400000)
- Launch: 23 November 2023; months from launch to the outcome: 23
- Team: 2 co-founders; 5 people listed by a directory in 2026; headcount at the figure not stated in any source read
- Outside funding: US$660K pre-seed, reported by one source
- Main factor: Focus
- Meters, 1 to 4: evidence strength 1, copyable pattern 2, capital needed 2
- Tags: AI-native, Small team, Funded, Running, B2B

Links:

- Visit runbear.io: https://runbear.io/
- Product Hunt: https://www.producthunt.com/products/runbear

**The product.** A company writes instructions for an AI agent, connects its documents and tools (Notion, Google Drive, HubSpot, Zendesk, Jira), and places the agent in Slack or Microsoft Teams with a name such as @Support or @SalesOps. Team members mention the agent in a channel; it answers from the connected knowledge, drafts replies and briefs, and updates records after a person approves. No code is needed.

**Why it worked.** The first version was an AI copilot for DevOps engineers, who could build their own agents. The founders moved the product to non-technical teams that answer the same questions all day, started every customer with one small use (repeated questions in Slack), and added a higher-priced plan when power users asked for more agents. The founder said in an interview of 17 October 2025, as quoted by SoFarBot, that this took the company from US$100,000 to US$400,000 in annual recurring revenue.

Evidence grades: A = disclosed by a public company, a filing or a payment processor, or verified by one; B = two or more independent reports agree, or a journalist confirmed it; C = a founder claim that no one else verified; D = reported, not verified: one founder claim or a third-party estimate with no second source; calc = computed on this page from graded facts.

## Summary

Runbear is a no-code platform for AI agents that live inside team chat. No-code means the user configures the software through forms and plain-language instructions instead of writing a program. An AI agent here is a program built on a large language model that reads a message, decides whether to answer, searches connected documents, calls other software tools and posts a reply. A team installs Runbear in Slack or Microsoft Teams, connects sources such as Notion, Google Drive, HubSpot and Zendesk, and then mentions the agent in a channel as it would mention a colleague.

Snow Lee and Liam Hwang, two software founders from South Korea who had worked together for 16 years, started the company in 2023 and launched the first Slack product, then called PlugBear, on Product Hunt on 23 November 2023. Three permitted secondary sources quote the founder's own figures: Startup Founder Stories and SoFarBot report that the founder stated US$100,000 in annual recurring revenue (ARR, the current monthly subscription revenue multiplied by twelve) in September 2024, followed by a plateau, and that in an Indie Hackers interview of 17 October 2025 he stated US$400,000 in ARR. Startup Founder Stories also reports that the founder said he raised US$660,000 in pre-seed funding.

Every money figure on this page is a founder statement quoted by a third party, or a directory estimate, so the page carries the grade D badge "Reported, not verified". The founder's own posts are on Medium, which this site's research rules do not read, and no payment processor has confirmed any figure. The page also shows four earlier companies that sold a chat bot or a chat platform to businesses and closed.

## Pricing

Runbear prices read on the company's pricing page on 23 October 2026 (grade C); the yearly column is computed from the page's "Yearly -20%" toggle

| Plan | Per month, billed yearly (US$) | Per month, billed monthly (US$) | Limit |
|---|---|---|---|
| Team | 79.20 | 99 | 20,000 credits a month, 5 builders, Slack, Discord and Teams |
| Business | 319.20 | 399 | 100,000 credits a month, 10 builders, HubSpot and Zendesk, bring your own key, SAML single sign-on, SOC 2 report |
| Enterprise | custom | custom | custom credits and builders, SCIM, HIPAA agreement on request |

Plans are priced by usage, not by seat. Each plan includes a monthly number of credits and a number of builders, the people who create and manage agents. A credit reflects the model work of a task: a quick knowledge-base answer is about 5 credits, a workflow with a few tool calls 20 to 50, a multi-page research task 300 to 1,500. Unused monthly credits do not roll over; pay-as-you-go buys more at the same rate. Business and Enterprise can bring their own model API key, and credits then become a small processing fee. The page shows only the monthly list prices and a 20% yearly discount; the yearly figures here are those prices times 0.8.

## How much revenue does Runbear make? The numbers, graded

A founder-reported US$400,000 in annual recurring revenue in October 2025, up from a founder-reported US$100,000 in September 2024, all quoted by secondary sources

| Fact | Value | Grade | Note | Sources |
|---|---|---|---|---|
| Annual recurring revenue, October 2025 | US$400,000 | D | Founder statement in an Indie Hackers interview of 17 October 2025, as quoted by SoFarBot (13 September 2026), Startup Founder Stories and a note.com article (16 February 2026). One founder claim with one date; no payment-processor confirmation. Startup Founder Stories adds that a December 2025 year-in-review post confirmed it, which this page could not read. | [1, 2, 3] |
| Monthly equivalent of the October 2025 figure | about US$33,333 a month | calc | Calculated: US$400,000 divided by 12. SoFarBot rounds this to US$33,000 a month and notes that it is an annualized figure, not cash received in one month. | [1] |
| Annual recurring revenue, September 2024 | US$100,000, then a plateau | C | Startup Founder Stories gives US$100,000 ARR in September 2024; SoFarBot summarizes a founder-written Reddit post that reported US$100,000 ARR after about nine months and then a plateau; Starter Story's page of 12 December 2024 carries the same figure in its title. Three permitted sources agree on the period. | [1, 2, 4] |
| Monthly revenue, December 2024 | US$8,300, profitable | D | Starter Story page dated 12 December 2024. The page does not say how the figure was obtained; it is consistent with US$100,000 a year. The page wrongly lists one founder. | [4] |
| Higher-priced Business plan | more than 20 customers, none churned by the interview date | D | Founder statement in the October 2025 interview, quoted by SoFarBot and the note.com article. SoFarBot dates the plan to July 2025; the note.com article says the single-plan period ended in July 2024. The two summaries disagree on the year. | [1, 3] |
| Price per user, December 2024 | US$29 a month with a free trial | D | Starter Story. Billing moved from messages to credits on 23 November 2025, according to the company's changelog, and the plans read in October 2026 are US$99 and US$399 a month. | [4, 8, 5] |
| Pre-seed funding | US$660,000 | D | Startup Founder Stories: "He received $660K in pre-seed funding." No investor, date or second source was found. | [2] |
| Team | 2 co-founders; 5 people listed in 2026 | D | The Org lists five people under the two founders in 2026 and the band "Employees 1-10", marked unverified. The co-founders' roles are on the company's author pages. No source read states the headcount at the October 2025 figure. | [13, 14] |
| Customers, October 2026 | 422 teams with 20 or more agent conversations in the past 12 months | C | Company home page, "Runbear production data, as of Oct 9, 2026". The page also states 1,200+ agents that did work and 35,000+ conversations handled in the last 28 days, on paid workspaces. | [6] |
| Named customers | Matillion, Fireflies.ai, BCG, Minted, Mercor, Tabs, TaxDome, Peak Ace, Storykit, SYPartners | C | Logos on the company's home and pricing pages. Case studies name Aloware, Peek.us, LaserAway, Power Real Estate Group and others. | [6, 16] |
| Time from the Product Hunt launch to the October 2025 figure | 23 months | calc | Calculated: 23 November 2023 to 17 October 2025. | [11, 1] |
| Product Hunt rating | 5.0 from 8 reviews, three launches | C | Product page: PlugBear launched 23 November 2023 and 7 May 2024, Runbear launched 24 June 2025. | [11] |
| Security audit | SOC 2 Type II, announced 25 July 2024 | C | Company blog. The changelog of 17 January 2024 said the audit was under way. | [10, 7] |
| Advertising cost per click | cut from US$1.06 to US$0.15 | D | Starter Story's account; it says early paid marketing did not work until the cost per click was lowered. | [4] |

How to read this table: a grade D value comes from one founder statement quoted by a newsletter or from a directory, with no second source. A grade C value is a founder figure that two permitted sources give for the same period, or a statement on the company's own pages. A "calc" value was computed on this page from the rows above it. No value is grade A or B.

## When did Runbear launch, and how did it grow? Timeline

From open-source developer tools in mid-2023 to a reported US$400,000 in annual recurring revenue in October 2025, 23 months after the first Product Hunt launch. Each entry carries its grade.

- **2023** (grade D [3, 4]): Snow Lee, a Seoul National University graduate who had founded and sold two companies and had been chief product officer at Buzzvil, starts Runbear with Liam Hwang, his co-founder of 16 years. Starter Story gives 2023 as the year started.
- **July to August 2023** (grade C [12]): The company publishes LangBear, an open-source prompt admin for LangChain, and awesome-runbook, a list of runbook resources, on Hacker News.
- **23 November 2023** (grade C [11, 2]): PlugBear, "Connect GPTs/Assistant to Slack and Discord", launches on Product Hunt. The first paying customer arrives in November 2023, according to Startup Founder Stories.
- **17 January 2024** (grade C [7]): Changelog: a /plugbear slash command for Slack, a SOC 2 Type II audit under way, and a Microsoft Teams app under review.
- **7 May 2024** (grade C [11]): Second Product Hunt launch: "Custom GPT for your Slack, Teams, HubSpot, Zendesk and more".
- **25 July 2024** (grade C [10]): SOC 2 Type II certification announced on the company blog.
- **28 August 2024** (grade C [9]): Changelog: new interface, "LLM Apps" renamed "Assistants", Google Gemini models added next to OpenAI.
- **September 2024** (grade C [1, 2]): The founder reports US$100,000 in annual recurring revenue, then a plateau, in a Reddit post summarized by SoFarBot; Startup Founder Stories dates the milestone to September 2024, one year after the first customer.
- **12 December 2024** (grade D [4]): Starter Story publishes a page stating US$8,300 in monthly revenue, profitable, at US$29 per user per month.
- **24 June 2025** (grade C [11]): Third Product Hunt launch under the name Runbear: "Your best new hire, but AI, in Slack".
- **July 2025** (grade D [1, 3]): A higher-priced Business plan replaces the single plan with an agent limit, after customers kept asking for more agents; more than 20 customers take it and none churns by October. SoFarBot dates this to July 2025; the note.com summary says July 2024.
- **17 October 2025** (grade D [1, 2, 3]): In an Indie Hackers interview the founder states US$400,000 in annual recurring revenue and describes the move from DevOps engineers to non-technical teams.
- **23 November 2025** (grade C [8]): Billing moves from a count of messages to credits, with pay-as-you-go top-ups.
- **9 October 2026** (grade C [6]): The home page states 422 customer workspaces with 20 or more agent conversations in the past year, and 35,000+ conversations handled in the last four weeks.
- **13 October 2026** (grade C [17]): Changelog: agents move to Claude 5.5 models, GPT-6.1 Sol and Gemini 3.8 Flash; an activity feed shows each agent's tool calls.

## Why did Runbear work?

### Stop selling to people who can build it themselves

The first product was a copilot for DevOps engineers. Engineers can connect a model to Slack on their own, so the customers were also the competitors. The founder told Indie Hackers, as quoted by the note.com article, that the people who used Runbear most were team leaders, customer-success staff, account managers and operations people, who knew the outcome they wanted and did not care how the model worked.

### One narrow first use

Almost every customer began with the same job: answering the questions that colleagues ask again and again in Slack. It is low-risk, read-only and easy to judge. Once a team trusted the agent with answers, it let the agent draft replies and update records in HubSpot or Zendesk. The founder said that if he started again he would find this killer use case before adding features.

### Let the limit requests set the price

Until the Business plan existed there was one subscription whose only limit was the number of agents. When customers kept asking for the limit to be lifted, the founders did not lift it; they added a plan with more support at a higher price. SoFarBot quotes the founder that more than 20 customers took it and none had churned by the interview date, which raised revenue per account without needing the same number of new customers.

### Live inside the tool the customer already opens

The agent is a Slack or Teams member with a name. There is no new login and no new dashboard for the people who ask questions; only the builders use Runbear's web app. The Slack Marketplace listing says the app runs in Slack's app container on paid Slack plans. Per-user authorization, approval steps, an audit log and a SOC 2 Type II report answer the objections of larger companies.

## How it was built

The stack below comes from the founder's interview as summarized by the note.com article, from the company's changelog and from the Starter Story page. All of it is grade D or C.

- TypeScript, hosted on Vercel, with Supabase as the database and backend service, according to the note.com summary of the founder's interview.
- Claude models for the core answer generation; OpenAI models for evaluation, knowledge extraction and compressing responses; Gemini and Perplexity available to customers who prefer them. The changelog of October 2026 lists Claude Opus, Sonnet and Haiku 5.5, GPT-6.1 Sol and Gemini 3.8 Flash, and a Claude Agent SDK runtime.
- Early versions wrapped the OpenAI Assistants API: the changelog of April 2024 moved customers to Assistants v2, and the January 2024 changelog imported OpenAI GPTs. Google Gemini was added in August 2024.
- Starter Story says the first version was built in seven days from existing APIs and launched on Product Hunt to collect feedback.
- Knowledge sources sync from Notion, Confluence, Google Drive, SharePoint, Slack and uploaded files; tools connect through MCP and 2,000+ integrations.
- SOC 2 Type II audit from January 2024, certified in July 2024; SAML single sign-on, SCIM and a HIPAA agreement on the upper plans.
- Marketing: use-case guides and customer stories on the company blog, creator collaborations on YouTube, X and LinkedIn, and search engine optimization for keywords about AI agents in chat channels. Starter Story adds Medium, Reddit and X posts and a paid-ads test whose cost per click was cut from US$1.06 to US$0.15.

Sources: [3, 17, 9, 7, 4, 5, 10]

## Who tried this before Runbear, and why did they fail?

Four companies in the Loot Drop database sold a bot or a chat platform to businesses before large language models existed, and closed. The values below are from that database and are not independently verified.

The failed startups come from Loot Drop (https://www.loot-drop.io/), a free database of dead startups, retold here in our own words; Loot Drop describes its entries as AI-assisted summaries of public sources, so they carry no evidence grade.

### Niles (2017 to 2020; funding US$2.8M; cause: Competition)

- Promise: An AI chatbot inside Slack that answers questions from the company's knowledge.
- Why it died: Its answers depended on 2017 language technology. Slack added its own search, and knowledge tools such as Guru and Confluence grew their features. Being only a Slack add-on left no room to move.
- Record: https://www.loot-drop.io/startup/1031-niles

### Answerly (2013 to 2016; funding US$550K; cause: Product/Tech Failure)

- Promise: FAQ bots that deflect support tickets so humans handle only the hard cases.
- Why it died: Natural language processing of the time could not handle real customer questions. Each customer needed months of manual FAQ writing and decision trees before the bot worked.
- Record: https://www.loot-drop.io/startup/2168-answerly

### Moped (2012 to 2017; funding US$1M; cause: Competition)

- Promise: A real-time team messaging platform with integrations, similar to Slack.
- Why it died: Slack grew faster with a better product and an ecosystem of integrations. Moped never differentiated and lost investor confidence.
- Record: https://www.loot-drop.io/startup/353-moped

### Vanti (2019 to 2024; funding US$18M; cause: No Market Need)

- Promise: Conversational AI chatbots that guide shoppers through product discovery and purchase.
- Why it died: Brands ran pilots and churned because chat-driven shopping did not beat the normal web shop on conversion. High acquisition cost and integration work killed the unit economics.
- Record: https://www.loot-drop.io/startup/2508-vanti

## What changed

Niles is the closest predecessor: a bot in Slack that answers questions from company knowledge. It died because the language technology of 2017 gave poor answers and because Slack and the knowledge tools added the feature themselves. Answerly died for the same first reason: each customer had to write FAQs and decision trees for months before the bot was useful.

Runbear builds on models that it does not have to train, so the answer quality that Niles and Answerly lacked arrives from Anthropic, OpenAI and Google with every model release, and the set-up that took Answerly months is a connection to Notion or Google Drive. The platform risk that killed Niles has not gone away: Slack sells its own AI, and the model makers sell team plans. Runbear's answer is to support every model and every chat tool, and to sell the parts the platforms do not: per-user permissions, approvals, audit logs and connections into the company's other software.

## Can one person copy Runbear now?

**Verdict: The method, yes; the market position, with care.**

The steps the founder described, as quoted by the note.com article and SoFarBot, are ordinary and repeatable: watch which customers get value, interview them, rank their problems by urgency, frequency, how many people have it and what it costs unsolved, then sell one narrow job first. The position is harder. A Slack agent platform competes with Slack, with the model makers and with every developer who can connect an API to a channel, and the founders had sixteen years of company building and, according to Startup Founder Stories, US$660,000 of pre-seed money before the figures on this page.

1. Pick a buyer who cannot build the agent alone and whose day is spent answering repeated questions in chat.
2. Start each customer with read-only question answering from one knowledge source, with citations, before giving the agent any write actions.
3. Put the agent where the questions already are, with no new login for the people who ask.
4. When customers ask you to lift a limit, add a higher plan instead of lifting it.
5. Get the security audit early; it is what larger customers ask for before the demo matters.

**What is still open.** Any internal workflow that runs through chat and a few business systems: IT help desk, HR questions, sales call follow-up, on-call engineering. Microsoft Teams in companies that do not use Slack.

## What could go wrong for Runbear?

- **What is not verified.** The headline figure, US$400,000 in annual recurring revenue, rests on one founder statement in an Indie Hackers interview of 17 October 2025, read here only through three summaries (SoFarBot, Startup Founder Stories and a note.com article). The earlier US$100,000 in September 2024 is also a founder figure, though three sources give it for the same period. The US$660,000 pre-seed round comes from one secondary source. No payment processor, investor or journalist has confirmed any of it. A reader could look for a TrustMRR or Stripe-verified profile, the founder's December 2025 year-in-review post on Medium, or a second interview with figures.
- **Headcount at the figure is not stated.** The only headcount source is The Org, an unverified directory that lists five people in 2026. With US$660,000 of pre-seed money the team at the October 2025 figure may have been larger than five, and no source read says.
- **The platform can close the window.** Slack sells its own AI, Microsoft sells Copilot inside Teams, and Anthropic, OpenAI and Google sell team plans with connectors. Niles, the closest predecessor, died when Slack added search. The Slack Marketplace listing notes that Runbear's agent experience in Slack's app container requires a paid Slack plan.
- **The two summaries disagree on a date.** SoFarBot says the single-plan period ended in July 2025; the note.com article says July 2024. The interview itself was not read.
- **Usage billing moves the model cost risk to the customer.** Since November 2025 a task costs credits in proportion to the model work it uses, and agents pause when credits run out. Heavy research or tool-calling agents can cost far more than the plan price suggests, and the company's own margin by workflow is not public.

## Sources

1. SoFarBot, 13 September 2026: How Runbear reached a reported $400K ARR: pivoting from DevOps copilots to AI teammates in Slack — https://www.sofarbot.com/stories/runbear-snow-lee-33000-monthly (grade D). Read in full. A source-linked summary of the Indie Hackers interview of 17 October 2025 and of a founder Reddit post; the source for the US$400,000 figure, the US$33,000 monthly equivalent, the Business plan and the pricing history.
2. Startup Founder Stories: Runbear Revenue: $400K ARR, $100K ARR in 1 year — https://startupfounderstories.com/stories/snow-lee-runbear-pivot-devops-to-400k-arr (grade D). Read in full. Summarizes the same interview; source for the October 2025 and September 2024 figures, the first customer in November 2023, the co-founders and the US$660,000 pre-seed round.
3. note.com, こうへい, 16 February 2026: The target for AI agents was wrong: a Korean founder's third company reaches ARR of 60 million yen (in Japanese) — https://note.com/aiedgerunner/n/nd5965c500c50 (grade D). Read in full. A Japanese summary of the Indie Hackers interview with direct quotations; source for the founders' background, the pivot, the ICP method, the stack and the growth channels.
4. Starter Story, 12 December 2024: How Snow W Lee Grew a No-Code AI Platform to $100K ARR — https://www.starterstory.com/runbear-breakdown (grade D). Read in full. Gives US$8,300 in monthly revenue, profitable, the seven-day first build, the US$29 price and the ad test. The page lists one founder, which other sources contradict.
5. Runbear pricing page, read 23 October 2026 — https://runbear.io/pricing (grade C). Read in full. Source for the plans, credits, builders and bring-your-own-key terms.
6. Runbear home page, read 23 October 2026 — https://runbear.io/ (grade C). Read in full. Source for the customer logos and the production figures dated 9 October 2026.
7. Runbear changelog, page 6 (November 2023 to January 2024) — https://runbear.io/changelog/page/6 (grade C). Read in full. Source for the PlugBear slash command, the SOC 2 audit under way and the Teams app in review.
8. Runbear changelog, 23 November 2025: Pricing Update, from messages to credits — https://runbear.io/changelog/2025-11-23-changelog (grade C). Read in full. Source for the move to credit billing and pay-as-you-go.
9. Runbear changelog, 28 August 2024: New Look, Google Gemini, and More — https://runbear.io/changelog/2024-08-28-revamp-ux (grade C). Read in full, with the 24 April 2024 entry on OpenAI Assistants v2 at https://runbear.io/changelog/2024-04-24-assistants-v2.
10. Runbear blog, 25 July 2024: Runbear is SOC 2 Type II Compliant! — https://runbear.io/posts/Runbear-is-SOC-2-Type-II-Compliant (grade C). Read in full. Company announcement of the certification.
11. Product Hunt: Runbear product page with three launches — https://www.producthunt.com/products/runbear (grade C). Read the launch list and the rating. Source for the launch dates of 23 November 2023, 7 May 2024 and 24 June 2025.
12. Hacker News search API (Algolia) for "runbear" — https://hn.algolia.com/api/v1/search?query=runbear&tags=story (grade C). Read the JSON result. Source for the LangBear and awesome-runbook posts of July and August 2023 and the Show HN posts of 2025.
13. The Org: Runbear, Inc. — https://theorg.com/org/runbear-inc (grade D). Read in full. An unverified crowd-sourced org chart listing the two founders, three other people and the band "Employees 1-10".
14. Runbear author pages: Snow Lee, CEO, and Liam Hwang, CTO — https://runbear.io/author/snow-lee (grade C). Read in full, with https://runbear.io/author/liam-hwang. Source for the founders' roles.
15. Slack Marketplace: Runbear — https://slack.com/marketplace/A065JPVAD3P-runbear (grade C). Read the listing. Source for the developer name, the connected tools and the paid-Slack-plan note.
16. Runbear case study: Meet Booker, Peek.us's AI Teammate in Slack — https://runbear.io/case-studies/Meet-Booker-Peekuss-AI-Teammate-in-Slack (grade C). Read in full, with the case-studies index at https://runbear.io/case-studies. Company-published customer results.
17. Runbear changelog, first page, 13 October 2026 entry — https://runbear.io/changelog (grade C). Read in full. Source for the current models and the Claude Agent SDK runtime.
18. Loot Drop database export, 1,749 failed startups — https://www.loot-drop.io/ (grade C). Local export. AI-written summaries, not independently verified. Source for the failed predecessors.