Skip to main content
Engineering work slips in quiet ways: an issue assigned months ago that nobody closed, action items from a planning meeting that never became tickets, a Sentry issue nobody picked up. Hive watches the tools your team works in, raises what’s gone stale, and turns loose ends into tickets and messages.

The problems Hive solves

  • Assigned work goes stale because priorities moved and the issue didn’t.
  • Meeting actions evaporate because nobody writes them into Jira.
  • Context lives in five tools (GitHub, Jira, Slack, Sentry, docs), so answering a simple status question takes a search in each one.
  • Internal tools and APIs don’t have a ready-made connector.

What Hive catches

The GitHub and Jira stale-work signals are switched on per workspace. Ask support if you’d like them. Hive reads at most 5,000 assigned items per run.

Meetings into tickets

The Meeting to Tasks template starts from a short form where you paste a meeting transcript, give the Jira project key for the new issues, and give an email address for the recap: Without an Approval step, a workflow carries out its action steps by itself, so each Jira issue is created as soon as the loop reaches it, whatever the Jira account’s level. To check the tasks first, add an Approval step between Extract tasks and For each task. See Workflow templates and What agents and workflows do by themselves.

Actions Hive can propose

In Ask and Signals, these engineering changes follow the Read, Draft or Act level of the account they use, and new accounts start on Act. On Draft, Hive proposes each one for your one-click approval. On Act, it makes the ones it can undo, such as closing or reopening a GitHub issue or adding a label, and tells you what it did; one it cannot undo, such as creating or commenting on an issue, still comes back as an approval card. An agent or workflow makes the ones selected for it by itself, unless the GitHub repository or Slack channel it acts on did not come from you: then it waits for your approval. Name a repository in an agent’s instructions by its GitHub link, and a Slack channel as #name. The actions are:
  • GitHub: create, comment on, update, close or reopen issues, and add labels.
  • Jira: create, comment on, assign, transition, update or label issues.
  • Sentry (through its MCP server or connector): assign, comment on, resolve or ignore issues.
  • Slack: post a message or reply in a thread.
Deleting a Jira issue is permanent, so it’s treated as irreversible. In Ask and Signals it always asks first, even on Act, and so does every tool on an MCP server. An agent or workflow you select it for deletes the issue by itself. Every change leaves a receipt in the audit log.

Merged-PR summaries in Slack

Hive can post a summary of merged pull requests from selected GitHub repositories to a Slack channel. It posts through the configuring member’s own Slack connection, so updates reach the team without anyone writing release notes by hand. Each summary covers up to 50 pull requests and says how many more come in later summaries, so nothing is skipped. Each pull request gets its own line, starting with its number, repository and author. It is plain text with no hidden links, @-mentions or link previews. See GitHub.

Bring your own systems

  • MCP servers. Connect any Model Context Protocol server over HTTPS, including the official Sentry, Linear and Atlassian (Confluence) servers. Tools are pinned when you connect, so a server can’t quietly change them later. Every tool is treated as high-risk until an operator says otherwise. See MCP servers.
  • Incoming webhooks. Send HMAC-signed JSON events from CI, monitoring or internal tools. See Incoming webhooks and Webhook signing.

Ask Hive for status

  • What’s assigned to me in Jira and GitHub that hasn’t moved in two weeks?
  • Summarise open pull requests in the payments repo.
  • Which Sentry issues in the web project are unresolved and assigned to nobody?
  • Draft a Slack update for #eng on what shipped this week.

For IT

  • Access reviews. Okta can suspend a user, which Hive treats as a revoke-access action (B4, highest risk). In Ask and Signals it always asks first, even on Act. Only an Owner or Admin can select it for an agent or workflow, which then suspends the user by itself. It can be reversed by unsuspending.
  • Workspace control. Admins decide which connectors members can use, see how many accounts of each app are connected (counts only), and can stop everything with the kill switch.
  • Security posture. Read the security overview and data protection.

Work and engineering integrations

GitHub, Jira, Linear, Sentry, Trello and more.

Developers

Webhooks, public forms and how Hive works.

Connect GitHub and Jira

Start a 7-day trial with no card required. Connect GitHub and Jira, set them to Read if Ask and Signals should only look, and see what’s stale.