# How to connect Mattermost to your company knowledge base > A step-by-step guide to connecting Mattermost to a self-hosted or air-gapped company knowledge base, so your team gets cited answers from real docs and code without leaving chat. Covers what you need, how the bridge works, deployment tiers, and how to set up @mentions, DMs, and auto-answering Q&A channels. - Date: 2026-06-16 - Tags: mattermost, knowledge-management, on-prem, air-gapped, integrations --- Mattermost is where your team talks. Connecting Falconer makes it where they get answers too, grounded in your real docs and code. To connect Mattermost to your company knowledge base with Falconer: deploy the Falconer agent in your environment, connect at least one knowledge source, run the lightweight Mattermost bridge alongside your instance, authenticate it with a Mattermost bot account, and configure the channels where it should respond. Falconer is a knowledge agent for engineering teams that unifies your docs, code, tickets, and chat into one knowledge graph and answers questions, with citations, for people and coding agents. This guide covers what you need, the steps to connect, what the integration does once it's live, and how it runs self-hosted or air-gapped. ## Key takeaways - Connecting Mattermost to Falconer runs a lightweight bridge alongside your instance that routes channel messages to Falconer's agent and posts cited answers back. - The bridge supports @mentions, DMs, auto-answering Q&A channels, threaded follow-ups, a remember command, and reaction feedback. - Setup requires a running Mattermost instance, admin access to create a bot account, and a Falconer deployment with a network path to Mattermost. - Falconer runs single-tenant inside your own GCP environment, and the full on-premises tier supports air-gapped operation with no outbound internet at runtime. - Answers are grounded in your connected docs and code and link to their sources, so every response is verifiable. - Falconer is SOC 2 Type II certified (achieved January 2026), encrypted in transit and at rest, and isolated inside your own VPC. The full security posture is published at [trust.falconer.com](https://trust.falconer.com). ## What you need - A running Mattermost instance (cloud or self-hosted) where you have admin access. - A Mattermost bot account and access token, so the bridge can read events and post replies. - A Falconer deployment: cloud, single-tenant on your own GCP project, or air-gapped on-premises. - At least one connected knowledge source (a repo, Notion, Linear, or similar) so the agent has content to answer from. - A network path between the bridge and your Mattermost instance. In air-gapped setups, both run inside your isolated environment. ## How to connect Mattermost to Falconer 1. Deploy Falconer in your environment. Choose cloud, single-tenant on your own GCP project, or air-gapped on-premises, depending on your compliance profile. 2. Connect at least one knowledge source. Point Falconer at a repo, Notion, Linear, or another integration so the knowledge graph has content to answer from. 3. Create a Mattermost bot account and generate an access token. 4. Run the Mattermost bridge alongside your instance, pointing it at your Mattermost URL and bot token. 5. Configure where the bot responds. Enable @mention answers, and set up a dedicated Q&A channel if you want every message answered automatically. 6. Test the connection. Mention the bot in a channel or post in the Q&A channel, and confirm you get an answer with citations back. For more on how the agent answers questions in chat, see [tools that answer questions from Slack using internal docs](https://falconer.com/guides/slack-question-answering-tools/) and [the best knowledge management tools with Slack integration for engineers](https://falconer.com/guides/slack-knowledge-management/). ## What you get Once connected, Falconer's full agent is available inside Mattermost: - **@mention responses.** Mention the bot in any channel and get an answer with citations and document references. - **Direct messages.** DM the bot for the same experience, privately. - **Q&A channels.** Configure a channel so every message is answered automatically, no mention needed. - **Threaded follow-ups.** The agent maintains context across replies in a thread. - **Remember command.** Store organizational knowledge for future answers. - **Reaction feedback.** Thumbs up or down on answers to improve quality. ![](https://falconer.com/api/file/s3/images/1781642351148-rgtu7o.png) ## How the connection works A lightweight bridge runs alongside your Mattermost instance. It listens to Mattermost's WebSocket events, forwards relevant messages to Falconer's agent, and posts the response back through the Mattermost REST API. The agent behind it is the same full Falcon agent you get in the Falconer web app. ## Built for engineering teams Engineers feel knowledge decay hardest. Code changes daily, docs rot, and decisions get buried in chat. On a self-hosted Mattermost stack, the usual cloud fix-it tools aren't an option, so the burden lands on whoever shipped the code last. Falconer takes that on: - **Docs that update from your PRs.** When a pull request merges, Falconer checks whether docs have drifted from the code and proposes updates. - **Codebase-aware answers.** Ask how something actually works and get an answer grounded in the real implementation, not a stale wiki page. - **Context for coding agents.** Feed accurate, grounded context to Claude, Cursor, and CLI agents over MCP. - **One place to search.** A single knowledge graph spans docs, code, tickets, and chat, so one answer can draw on a PR, a Linear issue, and a thread at once. See [how to build a company brain that connects your entire engineering stack](https://falconer.com/guides/company-brain-engineering-stack/) and [Falconer MCP](https://falconer.com/guides/falconer-mcp/) for how the knowledge graph feeds both people and coding agents. ## Self-hosted by default Because many Mattermost teams run self-hosted for compliance reasons, Falconer connects the same way. It runs as a single-tenant deployment inside your own GCP environment, with all container images baked in at build time so no external registry is needed at runtime. If you need full isolation, the full on-premises tier supports air-gapped deployments with no outbound internet access at runtime. All container images and model weights are pre-baked, so nothing phones home. Either way, Falconer is SOC 2 Type II certified (achieved January 2026), encrypted in transit and at rest, and isolated inside your own VPC, with time-limited, IP-restricted access and full audit logging. The complete security posture is at [trust.falconer.com](https://trust.falconer.com). ![](https://falconer.com/api/file/s3/images/1781642447140-cps1e.png) ### Two deployment tiers | Dimension | Managed on-premises | Full on-premises | | --- | --- | --- | | Where it runs | Your GCP project | Anywhere you can stand up the image, including air-gapped | | Who operates it | Falconer | Your team | | Upgrades | Falconer-pushed | Customer-controlled | | Air-gap support | No (GCP-connected) | Yes (zero outbound internet at runtime) | | Best for | Regulated teams wanting managed ops | Air-gapped, FedRAMP, ITAR, CMMC L3 | | Typical setup | 1-2 weeks | 2-4 weeks | For self-hosted and regulated-industry deployments, see [the best self-hosted Notion alternative for engineering teams](https://falconer.com/guides/self-hosted-notion-alternatives/) and [documentation platforms for AI coding assistants in defense tech startups](https://falconer.com/guides/defense-tech-documentation-platforms/). ## Falconer at a glance | Feature | Description | | --- | --- | | Knowledge graph | Unifies docs, code, and chat into one source of truth | | Auto-update from PRs | Docs refresh automatically as code changes | | Codebase awareness | Answers reflect your actual implementation | | Search | Combined semantic and keyword search | | Integrations | GitHub, Mattermost, Linear, Slack, Zendesk, and more | | Mattermost bridge | Full agent inside your channels and DMs | | Deployment | Cloud, on-prem, or air-gapped | ## Getting started 1. Connect one source first: a repo, Slack, or Linear. Falconer starts building your knowledge graph immediately. 2. Ask your first question right away. The graph gets richer as you add sources. 3. Run the Mattermost bridge and configure a Q&A channel so the agent answers automatically. Smaller teams layer Falconer over existing tools through unified search. Larger teams migrate fully, consolidating docs into one source of truth. See [how to consolidate documentation into one source of truth](https://falconer.com/guides/consolidate-documentation/) for both paths. ## FAQ ### How do I connect Mattermost to a knowledge base? Run Falconer's lightweight bridge alongside your instance. It connects to Mattermost's WebSocket API, forwards relevant messages to Falconer's agent, and posts cited answers back through the REST API. Point it at your Mattermost URL and a bot token, configure which channels it watches, and you get @mentions, DMs, and auto-answering Q&A channels. No Falconer-side infrastructure changes are needed. ### Can the Mattermost bot answer from our internal docs and code? Yes. The bot is the full Falconer agent, so answers are grounded in your connected docs, code, tickets, and chat, with citations that link back to the source. Because docs auto-update from your PRs, answers reflect what actually shipped rather than a stale wiki page. ### Does it work with a self-hosted Mattermost instance? Yes. Most Mattermost teams run self-hosted for compliance, and Falconer connects the same way. It runs single-tenant inside your own GCP environment, or fully air-gapped with no outbound internet at runtime. The bridge and your Mattermost instance both sit inside your perimeter, so no data leaves your network. ### Can a channel auto-answer questions without mentioning the bot? Yes. Configure a channel as a Q&A channel and Falconer answers every message automatically, no @mention required. It keeps context across threaded follow-ups, so people can ask clarifying questions in the same thread and get answers that build on what came before. ### What do I need before connecting Mattermost? A running Mattermost instance with admin access, a bot account and access token, a Falconer deployment (cloud, single-tenant GCP, or air-gapped on-premises), and at least one connected knowledge source so the agent has content to answer from. You also need a network path between the bridge and Mattermost; in air-gapped setups, both run inside your isolated environment.