MCP tools

The EkLine MCP server offers these tools. To connect a client, see Connect an MCP client.

Server

Item	Value
URL	https://ekline.io/api/mcp
Transport	Streamable HTTP
Authentication	An Authorization: Bearer <key> header with a personal API key. No OAuth.
Key format	ek_<id>_<secret>
Key scope	One member in one organization. Membership is checked on every call.
Missing, revoked or unknown key	The server answers with no tools rather than an error.

Tools

Tool	What it does	Main parameters
create_session	Starts a Docs Agent or Marketing Agent session and returns its id and a link to follow it. Returns once the session starts, not when it finishes.	prompt (required): the task. agent: docs (default) or marketing.
get_session_status	Reports a session’s status and up to six of its most recent messages.	sessionId (required)
send_session_message	Sends a correction, extra requirement or reference to a session that is running. Returns once the agent has the message.	sessionId and prompt (both required)
list_sessions	Lists your organization’s sessions: what each was asked and when it was last active. A pasted pull request, merge request, Jira or session URL finds the sessions related to it. Does not report whether a session is still running.	query: text to search for, or a URL. agent: docs, marketing or all (default). limit: 1 to 100 (default 20).
search_ekline_docs	Answers a question about EkLine from the published docs, with up to four sections and the URL of the page each came from.	query (required)
get_company_profile	Reads your organization’s company profile, including its named personas, competitors, brand terms and documentation sites.	None
update_company_profile	Replaces the profile fields you give, with no review step, and reports what each held before. Cannot clear a field or change named personas, competitors, brand terms or documentation sites.	At least one of companyProfile, keyFeatures, audience, competitors, toneAndVoice, visualStyle, additionalContext, each 1 to 10,000 characters
create_rules	Creates a Docs Agent rule, with no review step. Refuses a name your organization already uses. Cannot change, turn off or delete a rule.	name (required, up to 200 characters), content (required, up to 20,000 characters), ruleType: general (default) or review, paths: up to 20 file globs

Session status values

get_session_status returns one of these.

Status	Meaning
working	The agent is still running the session.
completed	The session finished.
failed	The session ran into a problem, or has had no activity for more than 15 minutes without finishing.
cancelled	The session was stopped.

Rule types

ruleType	Applies
general	On every task the agent works on
review	Before the agent opens or updates a documentation pull request

Journeys tools

Tool	What it does	Main parameters
list_journeys	Lists your organization’s journeys: each one’s id, title, kind, whether it is a draft or published, and how many reports it has.	None
get_journey	Reads one journey: what it asks, as which persona, its checks, which agents run it and how many times, and a summary of its latest report.	journeyId
get_journey_report	Reads one report: how each agent did on each check, the report’s findings, and an excerpt of each answer with the sources it cited.	journeyId, reportId
get_journey_run	Reads one run in full: every answer, every source cited, and each check result with its reason.	journeyId, reportId, runId
create_journey	Creates a draft journey. It does not run until you call run_journey.	sourcePrompt, title, kind (research or task), definition (1 to 20 turns, a persona, up to 50 checks), runConfig (agentIds and runCount)
update_journey	Changes a journey that has not run yet. Each field you give replaces the stored one in full.	journeyId, plus at least one of the create_journey fields
run_journey	Queues a run. It sends real requests to the AI agents and writes a report, usually within a few minutes. Returns once the run is queued.	journeyId

runConfig field	Values
agentIds	One or more of chatgpt, claude, perplexity, gemini, claude-code, codex. A task journey runs only on claude-code and codex.
runCount	How many times each agent runs the journey, 1 to 25
webSearch	Optional. On by default.
agentModels	Optional. Each agent uses its default model when left out.