Integration

Roe AI integration

Roe AI runs AI agents over unstructured data (documents, web pages, images) and a SQL workspace. List, create, version and duplicate agents; run them synchronously, as async jobs or as a batch, and read job status, results and artifacts; run SQL queries; manage datasets and upload or download files; manage policies and their versions; and manage webhooks that fire when linked agents finish a job. Every agent run, and any SQL that runs an agent, starts a job billed to your Roe credits. Not exposed: connections and connectors (they hold third-party credentials), users and organization administration, schedules, workflows, worksheets, knowledge bases, evaluations, tables and the platform's internal routes. Authentication: an organization API key from Settings in the Roe platform, sent as a Bearer token, plus your organization UUID, sent as the organization_id query parameter on every call.

Connect Roe AI How integrations work

Ask it in chat

Say what should happen.
The workflow builds itself.

When this happens, do that in Roe AI.Built from the 51 Roe AI tools below, on the connection you add once.

What you can do with Roe AI

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.

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

Agent, 8 operations

create_agentCreate agentPOSTCreates an agent and its first version on a given engine. Read list_engine_types for engine_class_id values and their input and config schemas, and list_supported_models for engine_config.model.
delete_agentDelete agentDELETEDeletes an agent. Returns no content.
get_agentGet agentGETReturns one agent: name, engine, current version id, cache settings, tags, creator and job statistics.
link_agents_to_webhookLink agents to webhookPOSTLinks one or more agents to a webhook, so every job those agents finish is POSTed to the webhook's url.
list_agentsList agentsGETLists the organization's agents (paginated): id, name, engine, current version, tags and job statistics. Filter by name, engine, creator, tags, dates or job counts.
list_webhook_agentsList webhook agentsGETLists the agents linked to a webhook; each entry's id is the link id unlink_agent_from_webhook takes.
unlink_agent_from_webhookUnlink agent from webhookDELETERemoves one agent's link from a webhook; the webhook stops receiving that agent's jobs. Returns no content.
update_agentUpdate agentPUTChanges an agent's name or cache settings; send only the fields to change. To change its prompt, inputs or engine configuration, create a new version with create_agent_version.

Version, 7 operations

create_agent_versionCreate agent versionPOSTCreates a new version of an agent (new inputs, engine configuration or post-actions). Read get_current_agent_version first and send back what should stay.
create_policy_versionCreate policy versionPOSTCreates a new version of a policy from the given content; it becomes the policy's current version and earlier versions are kept.
get_agent_versionGet agent versionGETReturns one version of an agent: its input definitions (the keys run_agent_version expects), engine configuration and post-actions.
get_policy_versionGet policy versionGETReturns one version of a policy: its content with the generated category and rule ids, author and parent policy.
list_agent_versionsList agent versionsGETLists every version of an agent with its input definitions, engine configuration and connector post-actions.
list_policy_versionsList policy versionsGETLists every version of a policy (paginated), with its content, author and the version it was derived from.
run_agent_versionRun agent versionPOSTRuns one specific version of an agent on the given inputs and waits for its outputs (a billed job). For long jobs use run_agent_version_async.

Policy, 5 operations

create_policyCreate policyPOSTCreates a policy (the rules a policy-aware agent checks against) with its first version, which becomes the current version.
delete_policyDelete policyDELETEDeletes a policy. Roe refuses it while any agent still uses the policy. Returns no content.
get_policyGet policyGETReturns one policy: name, description and current version id.
list_policiesList policiesGETLists the organization's policies (paginated): id, name, description and current version id.
update_policyUpdate policyPUTRenames a policy or changes its description. name is required: read get_policy first and send the current name back when only the description changes. Content changes go through create_policy_version.

Webhook, 5 operations

create_webhookCreate webhookPOSTCreates a webhook; link agents with link_agents_to_webhook so it receives a POST when their jobs finish (its id is in the answer, else read it with list_webhooks). The secret and headers you set are not returned.
delete_webhookDelete webhookDELETEDeletes a webhook. Roe refuses it with 400 while agents are still linked: unlink them first with unlink_agent_from_webhook.
get_webhookGet webhookGETReturns one webhook: name, url, events, active flag, delivery times and failure count. Its secret and custom headers are not returned.
list_webhooksList webhooksGETLists the organization's webhooks: name, url, events, active flag, delivery times and consecutive failure count. Secrets and custom headers are not returned.
update_webhookUpdate webhookPATCHChanges a webhook's name, url, events, active flag, secret, headers or alerts; send only what changes.

Dataset, 4 operations

create_datasetCreate datasetPOSTCreates an empty dataset (a folder for uploaded files) and returns its id.
delete_datasetDelete datasetDELETEPermanently deletes a dataset; its files are soft-deleted and can be restored later in Roe. Returns no content.
get_datasetGet datasetGETReturns one dataset: name, creator, creation date and organization.
list_datasetsList datasetsGETLists the organization's datasets (paginated): id, name, creator and creation date. Filter with search.

Result, 4 operations

get_job_artifactGet job artifactGETReturns the content of one artifact of a job (for example an evidence_data file named in get_job_result's outputs), the result field only.
get_job_resultGet job resultGETReturns a finished job's outputs (key, data type, value as a string, cost), its inputs and token counts. Output values that name artifact keys are read with get_job_artifact.
get_job_resultsGet job resultsPOSTReturns the outputs, status, cost, inputs and token counts of several jobs at once. Reads only, despite the POST.
get_sql_query_resultGet sql query resultGETReturns the rows of a query started with run_sql_query_async: column names and types and the result rows, with its status and any error.

Async, 3 operations

run_agent_asyncRun agent asyncPOSTStarts a billed job on the agent's current version and returns its job_id at once. Poll get_job_status until status is 3 (success), 4 (failure), 5 (cancelled) or 6 (cached), then call get_job_result.
run_agent_version_asyncRun agent version asyncPOSTStarts a billed job on one specific version of an agent and returns its job_id at once. Poll get_job_status, then call get_job_result.
run_sql_query_asyncRun sql query asyncPOSTStarts a SQL statement in the Roe SQL workspace and returns the query record at once; poll get_sql_query_status with its id, then read rows with get_sql_query_result. SQL that runs agents starts billed jobs.

Status, 3 operations

get_job_statusGet job statusGETReturns a job's status code: 0 pending, 1 started, 2 retry, 3 success, 4 failure, 5 cancelled, 6 cached (3 to 6 are final), with an error message on retry or failure.
get_job_statusesGet job statusesPOSTReturns the current status of several jobs at once (0 pending, 1 started, 2 retry, 3 success, 4 failure, 5 cancelled, 6 cached). Reads only, despite the POST.
get_sql_query_statusGet sql query statusGETReturns the status of a query started with run_sql_query_async, with its error message if it failed and when it finished.

File, 2 operations

delete_dataset_fileDelete dataset fileDELETEDeletes one file from a dataset. Returns no content.
list_dataset_filesList dataset filesGETLists the files in a dataset (paginated): id, name, size, content type, creator and metadata. Filter by name with search.

Current, 1 operation

get_current_agent_versionGet current agent versionGETReturns the agent's current version: its input definitions (the input keys run_agent expects), engine configuration and post-actions.

Download, 1 operation

download_fileDownload fileGETDownloads a stored file by its id and returns it as a file. For a PDF, page_range keeps only some pages.

Duplicate, 1 operation

duplicate_agentDuplicate agentPOSTCreates a new agent with the same attributes as the current version of this agent and returns that new agent's version.

Job, 1 operation

list_agent_jobsList agent jobsGETLists an agent's jobs (paginated): status, status history, inputs, metadata, duration, review and evaluation. Filter by status, version, dates or text.

Many, 1 operation

run_agent_batch_asyncRun agent batch asyncPOSTStarts one billed job per item of inputs on the agent's current version and returns the list of job ids. Read statuses with get_job_statuses and outputs with get_job_results. Metadata is ignored here.

Model, 1 operation

list_supported_modelsList supported modelsGETLists the model ids accepted in engine_config.model, with providers, input capabilities, context window and feature support. The list is the same for every organization.

Query, 1 operation

run_sql_queryRun sql queryPOSTRuns a SQL statement in the organization's Roe SQL workspace and waits for its rows. SQL can call agent functions such as run_agent, which start billed jobs.

Run, 1 operation

run_agentRun agentPOSTRuns an agent's current version on the given inputs and waits for its outputs (a billed job). For long jobs use run_agent_async, then get_job_status and get_job_result.

Type, 1 operation

list_engine_typesList engine typesGETLists the engine_class_id values create_agent accepts, with each engine's description, input schema and default engine_config.

Upload, 1 operation

upload_fileUpload filePOSTUploads a file to Roe and returns its id, name, size and dataset. Pass the id as a file input to run_agent or run_agent_async.

Use Roe AI from Claude Code, Cursor or Codex

Connect Roe AI in LiveContext, create an API key in Settings, then add the LiveContext MCP server to your client. Its Roe AI 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 Roe AI

Create your workspace
Sign up free. Nothing to install.
Connect Roe AI once
Add your Roe AI 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.