Connect one app this server uses. For an app that issues keys, says where to get one and the terminal command that stores it. For Google, once the user's own OAuth client is set, starts the browser sign-in and returns at once: the user approves in the browser and the next call works. To see which apps are connected, call connection_status. Never ask the user for a key in the chat.

See which apps this server is connected to, and how to connect each one that is not. Changes nothing.

List checks for a monitor.

Create a scheduled monitor for scrape, crawl, or search targets.

Create a calendar event; returns details of the event.

List events matching a given search filter.

Create a blank document with a title. Only the title is honoured — the document is created empty. To add content, call this and then documents_batch_update with the returned documentId.

Read a document's full structural content. The document ID is the long string in its URL, between '/d/' and '/edit'. Pass include_tabs_content=true for a document with several tabs, since the default response covers only the first.

Read a failing job's log. Returns the last 200 lines plus every line that looks like an error, with the line numbers and the log's real length — not the whole file, which is routinely megabytes. Use this after actions_list_jobs_for_run has named the job; try checks_list_annotations first, which is cheaper when the check reported the failure with a file and a line.

Get a run's total run time and its billable milliseconds per runner operating system. Billable minutes apply only to private repositories on GitHub-hosted runners. GitHub has this endpoint closing down.

List a run's jobs, each with its conclusion and its failing steps named. This is what turns 'the run failed' into 'the pytest step of the test job failed', usually without reading a log at all.

List the environments a run is waiting on for approval, and who can approve them. This is where a run sits when it is neither failing nor progressing.

List the runs of one workflow, most recent first — 'is this pipeline green', rather than 'is anything red'. The same single status filter as actions_list_workflow_runs, which matches a status or a conclusion.

List CI runs, most recent first. The single status filter matches either a run's status or its conclusion, so 'in_progress' and 'failure' are both valid there.

Approve or reject a run's pending deployments. Takes the environment ids from actions_list_pending_deployments. A comment is required by GitHub, not optional.

List a check run's annotations: each failure with the file, the line range, a level and a message, already extracted by whoever ran the check. The cheapest way to find out why a build is red — try this before downloading any log. GitHub returns at most 50 per page.

List code scanning alerts, each with the rule that fired and the file and line it fired on. Filter to one pull request with pr, or to a branch with ref. Needs GitHub Advanced Security on a private repository, which answers 403 without it.

List a repository's Dependabot alerts: which dependency, how severe, and the first version that fixes it. Several filters take a comma-separated list, so state='open,dismissed' is valid. Needs the security_events scope. This endpoint has no page parameter — raise per_page rather than paging.

Create a branch or tag pointing at an existing commit. The reference must be fully qualified, like 'refs/heads/my-branch'.

Get a reference and the commit SHA it points at — how you read a branch's current head before building on it. ref is heads/main or tags/v1.0, without the refs/ prefix.

Post a comment on an issue or pull request. Pull requests take their own number here — GitHub treats them as issues for commenting.

List issues in a repository. Note that GitHub counts pull requests as issues, so results include both; the ones that are pull requests are marked is_pull_request. Filter with state, labels, assignee or creator.

List your unread notifications — mentions, review requests, assignments — most recently updated first. This is how an agent finds out it is needed without polling repositories. reason says why each one is here; url is the thing it concerns. Note per_page maxes at 50 here, not 100.

Open a pull request from head into base. Provide title for a new pull request, or issue to convert an existing issue into one — exactly one of the two. Both branches must already exist.

Get a single pull request, including its mergeability and the counts of changed files, additions and deletions. These details are only present on this endpoint, not on pulls_list.

Get a pull request's whole change as a unified diff, in one call. The same endpoint as pulls_get asked for a different representation. Use it to read a change end to end; use pulls_list_files when the change is large enough to need paging, since this has no way to ask for less.

List pull requests. Filter by state, by base branch, or by head in the form 'user:branch'.

List the files a pull request changes, with per-file additions, deletions and patch text. A large pull request returns a lot of diff: keep per_page small and page through it.

List who has been asked to review and has not answered yet. A reviewer drops off this list once they submit, so an empty list means nobody is outstanding — not that nobody was asked.

Update a pull request's title, body, state or base branch. Closing one is state: closed; there is no REST way to change its draft status.

Get the latest published release — GitHub's definition, meaning the most recent release that is neither a draft nor a prerelease. Not simply the newest one.

Update a release; anything omitted is left unchanged. Publishing a draft is this call with draft: false.

Compare two refs and get what is in one and not the other: the status, how far ahead or behind, the commits and the changed files. basehead is one string, 'main...my-feature', with three dots. Across forks, prefix each side with its owner.

Create a file or replace an existing one, in a single commit. Write the content as plain text. To replace a file, read it first and pass its sha; without one GitHub treats this as a create and refuses if the path exists.

Get a repository's metadata, including its default branch and topics.

Read a file, or list a directory. Pass an empty path for the repository root. File contents come back decoded as text.

Get the repository's README, decoded to text. GitHub picks the preferred file itself across the spellings a repository might use, which saves guessing at repos_get_content.

List commits, newest first. Narrow with sha for a branch, path for one file's history, or since/until for a window.

List contributors, most commits first — who to ask about a repository. GitHub links only the first 500 author email addresses to accounts; the rest come back anonymous. An empty repository returns an empty object.

List a repository's languages with bytes of code each — the quickest way to find out what a repository is written in before reading any of it. Answers with an object keyed by language, so there is nothing to page.

Search code across GitHub. Only default branches are indexed and only files under 384 KB. The query needs a real search term, not just qualifiers. This endpoint allows ten requests per minute and returns at most 1,000 results.

Search issues and pull requests across GitHub with qualifiers, e.g. 'repo:owner/name is:issue is:open label:bug'. Use this when the repository is unknown or the filter is richer than issues_list_for_repo supports. Rate limited to 30 requests per minute.

Search repositories with qualifiers, e.g. 'topic:cli language:go stars:>500'. Rate limited to 30 requests per minute.

List secret scanning alerts: which pattern matched, which provider, and whether the credential is still valid. The detected secret itself is never returned by this tool — it is dropped from the response whatever the request asked for, because triage does not need it. Requires administrator access.

Appends values to a spreadsheet.

Sets values in a range of a spreadsheet.

Link an issue to something outside Linear. Linear deduplicates by URL: creating an attachment whose URL is already on the issue updates that one rather than adding a second.

Find which issues an external URL is attached to. This is the tool for 'is this pull request already linked to an issue?'.

Create an issue. team_id and title are required; everything else is optional. The UUIDs for team, assignee, state and labels come from teams_list, users_list and workflow_states_list — Linear does not accept names here. Set parent_id to create a sub-issue.

Get one issue with its description, comments, sub-issues, relations and attachments. Accepts either the issue's UUID or its human identifier, such as 'ENG-123'.

Link two issues as blocking, duplicate, related or similar. 'blocks' and 'duplicate' are directional: the issue in issue_id is the one doing the blocking, or the one that is the duplicate.

List issues, optionally filtered. Conditions on one filter object combine with AND. To find a team's open work, filter on team.key and state.type.

Search issues by text, across titles and descriptions. Set include_comments to search inside comments too. This is full-text search; to filter on fields such as state or assignee, use issues_list.

Create a page — as a subpage of another page, or as a row of a database by giving its data_source_id as the parent. Content comes as a markdown string Notion parses into blocks, or from a template: one or the other, never both.

Replace a page's content with Markdown, which Notion parses into blocks. This replaces the whole body, so read it first if you mean to edit rather than overwrite.

Send a message to a Slack channel, private group, or DM. Provide text for a plain message; set thread_ts to reply inside an existing thread.

Fetch recent messages from a Slack channel.

Open a direct message with one person, or a group message with up to eight. The channel id for a following chat_post_message comes back at channel.id. Opening the same conversation twice returns the existing one rather than making another.

Fetch a thread of messages. Pass the parent message's ts as the thread identifier.

List members of the workspace. Use this to resolve a person's name to the user ID that Slack mentions and filters expect.

Execute a real-time web search optimized for AI agents. Use when sources are unknown or current web context is needed. Prefer search_depth advanced with chunks_per_source 3 for stronger evidence per source.