Integration

Langdock integration

Langdock, the EU-hosted AI workspace of Langdock GmbH (Berlin). Chat with your Langdock agents, call OpenAI, Anthropic and Google Gemini models through provider-compatible endpoints, create embeddings and images, search and manage Knowledge bases, manage agents, Skills, prompts, scheduled tasks and workflows, and export usage reports. Model and agent usage counts toward the key's budget and the workspace spend limit. Not exposed: streaming, the Codestral route (Langdock says no Mistral model is available through the API), the admin APIs (user management, billing, audit logs, governance, agent disabling), the Integrations, Meetings and Compliance APIs, and the node payloads of workflow runs. Authentication: a Langdock personal or workspace API key sent as a Bearer token; each API family needs its scope on the key. Langdock Cloud only: dedicated deployments use their own host.

Connect Langdock How integrations work

Ask it in chat

Say what should happen.
The workflow builds itself.

When this happens, do that in Langdock.Built from the 70 Langdock tools below, on the connection you add once.

What you can do with Langdock

Each of these is one operation on one endpoint, available both as a workflow step and as a tool an AI agent can call on its own. Nothing bundles several operations into one action, so what a run did stays readable afterwards.

Each badge shows the default agent permission of an endpoint (allowed: an agent calls it without asking; ask first: you are asked before an agent calls it, when the agent or chat asks permission for sensitive actions; never: no agent may call it), which every user can change endpoint by endpoint in their credential settings.

Most endpoints run on the free plan. A few, mainly the ones that publish on your behalf, need a paid plan.

Scheduled task, 8 operations

create_scheduled_taskCreate scheduled taskPOSTAgents: ask firstCreates a scheduled task that sends a prompt to Langdock on a schedule (or only on demand with MANUAL). It starts active and runs at the next scheduled time with its agent or model. Returns the task.
delete_scheduled_taskDelete scheduled taskDELETEAgents: ask firstPermanently deletes a scheduled task (automationId from list_scheduled_tasks); it stops running.
get_scheduled_taskGet scheduled taskGETAgents: ask firstReturns one scheduled task (automationId from list_scheduled_tasks) with its schedule, prompt, agent, tags, active state and run count.
list_scheduled_tasksList scheduled tasksGETAgents: ask firstLists the scheduled tasks owned by this key's member or service account, newest first, with schedule, prompt, agent, active state and run count. Paginate with nextCursor.
pause_scheduled_taskPause scheduled taskPOSTAgents: ask firstPauses a scheduled task so it stops running on its schedule (active becomes false). Resume it with resume_scheduled_task.
resume_scheduled_taskResume scheduled taskPOSTAgents: ask firstResumes a paused scheduled task so it runs on its schedule again (active becomes true).
run_scheduled_taskRun scheduled taskPOSTAgents: ask firstStarts one run of a scheduled task now (also when paused, without resuming it). Returns runId as soon as the run is queued, without waiting for it.
update_scheduled_taskUpdate scheduled taskPATCHAgents: ask firstChanges a scheduled task's prompt, schedule, model, agent or tags (automationId from list_scheduled_tasks). Send at least one field; only those change, and the merged schedule must stay complete.

Knowledge file, 6 operations

delete_knowledge_fileDelete knowledge fileDELETEAgents: ask firstPermanently deletes a file and its embeddings from a Knowledge base; it disappears from search at once and cannot be recovered. Key needs the EDITOR role.
get_knowledge_fileGet knowledge fileGETAgents: ask firstReturns one file of a Knowledge base with its processing state: poll it after upload_knowledge_file until syncStatus is SYNCED (searchable) or a failure status (then syncMessage says why).
list_knowledge_filesList knowledge filesGETAgents: ask firstLists every file of a Knowledge base with name, type, processing state (syncStatus: SYNCED = searchable) and auto-generated summary.
replace_knowledge_fileReplace knowledge filePATCHAgents: ask firstReplaces a file of a Knowledge base with a new version (attachmentId from list_knowledge_files): the old file and its embeddings are removed and the new one is processed. Key needs the EDITOR role.
upload_knowledge_fileUpload knowledge filePOSTAgents: ask firstUploads a document to a Knowledge base (key needs the EDITOR role) and returns its attachment id. Processing for search continues afterwards: poll get_knowledge_file until syncStatus is SYNCED.
upload_knowledge_file_asyncUpload knowledge file asyncPOSTAgents: ask firstSame as upload_knowledge_file but answers as soon as the file is received (HTTP 202), for large files; processing continues in the background, so poll get_knowledge_file until syncStatus is SYNCED.

Skill, 6 operations

create_skillCreate skillPOSTAgents: ask firstCreates a Skill in the workspace from a name and instructions, optionally with a slug, a description of when it applies and integration ids. Returns the Skill with its id.
delete_skillDelete skillDELETEAgents: ask firstPermanently deletes a Skill from the workspace (skillId from list_skills).
get_skillGet skillGETAgents: ask firstReturns one Skill (skillId from list_skills) with its instructions and the list of its stored files (path, size, extension, sha256). Read a text file with get_skill_file.
import_skillImport skillPOSTAgents: ask firstCreates a Skill from a SKILL.md file or a zip archive. mode upsert instead overwrites the editable Skill that has the same slug.
list_skillsList skillsGETAgents: ask firstLists the Skills of the workspace (reusable instructions agents can apply), with slug, description and instructions. Filter by query or slug; paginate with nextCursor. Needs the Skills API scope.
update_skillUpdate skillPATCHAgents: ask firstUpdates a Skill (skillId from list_skills). Only the fields you send change; integrationIds replaces the whole list when sent.

Workflow, 6 operations

create_workflowCreate workflowPOSTAgents: ask firstCreates an inactive draft workflow (nothing runs until publish_workflow): give a name and either a starter trigger (initialTriggerKind) or a full graph (nodes and edges). shareWith shares it with people.
delete_workflowDelete workflowDELETEAgents: ask firstPermanently deletes a workflow with its draft and every published version; there is no confirmation step. Needs the Workflow Deletion API scope (workspace keys only) and editor access.
get_workflowGet workflowGETAgents: ask firstReturns one workflow (workflowId from list_workflows) with its draft graph (nodes, edges), published version, version history and execution limits. Secret-like fields are redacted by Langdock.
list_workflowsList workflowsGETAgents: ask firstLists the Langdock workflows visible to the key (workspace keys see all non-template workflows): name, status, owner, your access role and published version. No graph: use get_workflow. Paginate with nextCursor.
publish_workflowPublish workflowPOSTAgents: ask firstPublishes a workflow's draft as a new production version: it becomes ACTIVE and its scheduled and webhook triggers start running real executions. Requires bumpType.
update_workflowUpdate workflowPATCHAgents: ask firstUpdates a workflow's metadata, status or draft graph (workflowId from list_workflows). status INACTIVE pauses a live workflow; ACTIVE needs a published version. Graph edits apply to the draft only.

Agent, 5 operations

create_agentCreate agentPOSTAgents: ask firstCreates a Langdock agent in the workspace (as a draft until publish_agent publishes it). Requires name. The access field shares it with people and groups right away. Returns the agent with its id.
get_agentGet agentGETAgents: ask firstReturns one agent's settings (agentId from list_agents): instructions, model, capabilities, actions, form fields, attachments and Knowledge bases. Published version by default; version=draft returns the draft.
list_agentsList agentsGETAgents: ask firstLists the agents shared with this API key (not every agent of the workspace), with name, model, input type and owner (name and email: personal data). Paginate with nextCursor.
publish_agentPublish agentPOSTAgents: ask firstPublishes the current draft of an agent as a new version that every user of the agent gets (agentId from list_agents). The draft must differ from the latest published version.
update_agentUpdate agentPATCHAgents: ask firstUpdates a Langdock agent's draft (agentId from list_agents); publish_agent makes the change live. Only the fields you send change. access changes take effect at once and grant real access.

Prompt, 5 operations

create_promptCreate promptPOSTAgents: ask firstAdds a prompt to the prompt library. Put it in a folder (promptFolderId) or share it with the whole workspace (sharedWithWorkspace), not both; sharing makes it visible to every member.
delete_promptDelete promptDELETEAgents: ask firstPermanently deletes a prompt from the prompt library (promptId from list_prompts).
get_promptGet promptGETAgents: ask firstReturns one prompt of the prompt library (promptId from list_prompts) with its text, folder and sharing.
list_promptsList promptsGETAgents: ask firstLists the prompts of the prompt library you can access: title, prompt text, folder and whether it is shared with the workspace. Filter by query, folder or sharing; paginate with nextCursor.
update_promptUpdate promptPATCHAgents: ask firstUpdates a prompt (promptId from list_prompts). Only the fields you send change; sharedWithWorkspace true makes it visible to every member.

Prompt folder, 5 operations

create_prompt_folderCreate prompt folderPOSTAgents: ask firstCreates a prompt folder. Share it with the whole workspace (sharedWithWorkspace) or with one group (sharedWithGroupId), not both; sharing makes its prompts visible to those members.
delete_prompt_folderDelete prompt folderDELETEAgents: ask firstPermanently deletes a prompt folder (folderId from list_prompt_folders).
get_prompt_folderGet prompt folderGETAgents: ask firstReturns one prompt folder (folderId from list_prompt_folders) with its name and sharing.
list_prompt_foldersList prompt foldersGETAgents: ask firstLists the prompt folders you can access, with name and sharing (workspace or one group). Filter by query or sharing; paginate with nextCursor.
update_prompt_folderUpdate prompt folderPATCHAgents: ask firstRenames a prompt folder or changes its sharing (folderId from list_prompt_folders). Only the fields you send change.

Knowledge base, 4 operations

create_knowledge_baseCreate knowledge basePOSTAgents: ask firstCreates an empty Knowledge base owned by this workspace API key (the key gets the EDITOR role) and returns its id; then add files with upload_knowledge_file. Personal keys are refused. There is no delete route.
list_knowledge_basesList knowledge basesGETAgents: ask firstLists the Knowledge bases this workspace API key can use, with id, name, file count and the key's role (USER = Viewer, EDITOR = Editor). Personal keys get an empty list. Paginate with nextCursor.
search_knowledge_basesSearch knowledge basesPOSTAgents: ask firstSemantic search across every Knowledge base shared with this API key: returns the best-matching text chunk of each relevant document, with similarity (0 to 1), file id and file name.
update_knowledge_baseUpdate knowledge basePATCHAgents: ask firstRenames a Knowledge base or changes its description (the key needs the EDITOR role on it). Send at least one of name and description; leave one out to keep it.

Knowledge base access, 3 operations

grant_knowledge_base_accessGrant knowledge base accessPOSTAgents: ask firstGrants people or workspace API keys access to a Knowledge base: role USER (Viewer: search and read) or EDITOR (upload, delete, share, rename). Gives real access to its documents. Key needs EDITOR.
revoke_knowledge_base_accessRevoke knowledge base accessDELETEAgents: ask firstRemoves one person's or workspace API key's access to a Knowledge base. The owner's access cannot be revoked. Key needs the EDITOR role.
set_knowledge_base_accessSet knowledge base accessPATCHAgents: ask firstChanges the role of one person or workspace API key on a Knowledge base, creating the grant if it does not exist yet. Changes real access. The owner's access cannot be changed. Key needs EDITOR.

Attachment, 2 operations

delete_attachmentDelete attachmentDELETEAgents: ask firstDeletes an uploaded attachment (soft delete) by attachmentId. The attachment must have been created by the key owner or belong to a Knowledge base shared with the key.
upload_attachmentUpload attachmentPOSTAgents: ask firstUploads a file for agent conversations and returns attachmentId; put it in a message's metadata.attachments in chat_with_agent, or in attachments of create_agent. Needs the Knowledge Folder API scope.

Workflow run, 2 operations

export_workflow_runsExport workflow runsGETAgents: ask firstExports one row per node execution of a workflow's runs in a date range (from and to required), oldest first: run and node ids, types, statuses and times. Not paginated: at most 10,000 runs per call.
list_workflow_runsList workflow runsGETAgents: ask firstLists runs of a workflow (workflowId from list_workflows), newest first, with status, test or production mode, version and each node execution's status and duration. Needs editor access.

Agent conversation, 1 operation

chat_with_agentChat with agentPOSTAgents: ask firstSends messages to a Langdock agent (agentId from list_agents) or to a temporary agent (agent) and returns its reply in messages[].content. The agent may run its tools; usage counts toward the spend limit. No streaming; fails after 100 s (HTTP 524).

Agent model, 1 operation

list_agent_modelsList agent modelsGETAgents: ask firstLists the models available to agents in this workspace: the id to send as model to create_agent or in an inline agent, its region and whether it supports extended thinking.

Agent usage, 1 operation

export_agent_usageExport agent usagePOSTAgents: ask firstExports per-agent usage (messages, active users, trends) for a date range as a CSV file and returns its signed download URL. Read only; needs the Usage Export API scope.

Anthropic model, 1 operation

list_anthropic_modelsList anthropic modelsGETAgents: ask firstLists the Claude models available to the workspace in a region; pass an id as model to create_anthropic_message with the same region. Includes models that support extended thinking.

Api key usage, 1 operation

export_api_key_usageExport api key usagePOSTAgents: ask firstExports usage and cost per API key for a date range as a CSV file and returns its signed download URL. Read only; needs the Usage Export API scope.

Chat completion, 1 operation

create_chat_completionCreate chat completionPOSTAgents: ask firstGenerates a chat reply with an OpenAI or Azure OpenAI model of the workspace (model id from list_openai_models, same region). Counts toward the key's budget and the workspace spend limit. No streaming.

Embedding, 1 operation

create_embeddingCreate embeddingPOSTAgents: ask firstCreates embedding vectors for one text or a list of texts. Langdock supports only the text-embedding-ada-002 model here. Counts toward the key's budget and the workspace spend limit.

Gemini content, 1 operation

generate_gemini_contentGenerate gemini contentPOSTAgents: ask firstGenerates content with a Gemini model hosted in Google Vertex AI (model id from list_gemini_models, without the models/ prefix), in the Gemini generateContent format. Counts toward the workspace spend limit. No streaming.

Gemini model, 1 operation

list_gemini_modelsList gemini modelsGETAgents: ask firstLists the Gemini models available in a region (name like models/gemini-2.5-flash); drop the models/ prefix and pass the rest as model to generate_gemini_content.

Image, 1 operation

generate_imageGenerate imagePOSTAgents: ask firstGenerates one image from a prompt with an image model of the workspace and returns it as base64 (data[0].b64_json). Billed as image generation; times out after 60 s. One image per call, no URL output, no HD, no edits.

Message, 1 operation

create_anthropic_messageCreate anthropic messagePOSTAgents: ask firstGenerates the next assistant message with a Claude model of the workspace (model id from list_anthropic_models, same region), in the Anthropic Messages format. Counts toward the workspace spend limit. No streaming; times out after 105 s.

Model usage, 1 operation

export_model_usageExport model usagePOSTAgents: ask firstExports per-model usage (request counts, BYOK token consumption) for a date range as a CSV file and returns its signed download URL. Read only; needs the Usage Export API scope.

Openai model, 1 operation

list_openai_modelsList openai modelsGETAgents: ask firstLists the OpenAI-compatible chat models available to the workspace in a region; pass an id as model to create_chat_completion or create_response with the same region. Image models are not listed.

Project usage, 1 operation

export_project_usageExport project usagePOSTAgents: ask firstExports per-project activity and resource use for a date range as a CSV file and returns its signed download URL. Read only; needs the Usage Export API scope.

Response, 1 operation

create_responseCreate responsePOSTAgents: ask firstGenerates a reply in the stateless OpenAI Responses format (model id from list_openai_models, same region); send the whole conversation each time. Counts toward the key's budget and the workspace spend limit. No streaming.

Skill file, 1 operation

get_skill_fileGet skill fileGETAgents: ask firstReturns the content of one text file stored in a Skill (path as listed by get_skill), with size, extension and sha256. UTF-8 text files up to 512 KB only.

User usage, 1 operation

export_user_usageExport user usagePOSTAgents: ask firstExports per-user usage (messages, models, spend; names and emails, when the workspace allows them, are personal data) for a date range as a CSV file and returns its signed download URL. Read only; needs the Usage Export API scope.

Workflow usage, 1 operation

export_workflow_usageExport workflow usagePOSTAgents: ask firstExports per-workflow cost, run counts and monthly cap use for a date range as a CSV file and returns its signed download URL. Read only; needs the Usage Export API scope.

Use Langdock from Claude Code, Cursor or Codex

Connect Langdock in LiveContext, create an API key in Settings, then add the LiveContext MCP server to your client. Its Langdock operations become tools the client can call, next to every other integration you connected.

Claude Code

claude mcp add --transport http livecontext https://livecontext.ai/mcp --header "X-API-Key: YOUR_API_KEY"

Codex (~/.codex/config.toml)

[mcp_servers.livecontext]
url = "https://livecontext.ai/mcp"

[mcp_servers.livecontext.http_headers]
"X-API-Key" = "YOUR_API_KEY"

Cursor, Claude Desktop and other JSON configs

{
  "mcpServers": {
    "livecontext": {
      "type": "http",
      "url": "https://livecontext.ai/mcp",
      "headers": {
        "X-API-Key": "YOUR_API_KEY"
      }
    }
  }
}

Set up the MCP server step by step

How to connect Langdock

Create your workspace
Sign up free. Nothing to install.
Connect Langdock once
Add your Langdock connection in Settings. Every tool from this integration reuses it, and secrets are encrypted at rest and visible only to you.
Describe the job in chat
The workflow builds itself in front of you and picks the endpoints it needs. Drag the steps yourself on the canvas whenever you would rather be explicit.