Create Agent

Creates an agent for your workspace. A workspace can have multiple agents. The agent is created enabled (enabled: true) and starts monitoring immediately.

Credit usage

champion_tracking agents consume credits from your workspace balance each time a tracked champion is re-enriched — a recurring cost for as long as the agent stays enabled. Keyword agents (keyword_detection, keyword_post_search, job_posting_keyword_detection) consume credits for the keyword signals they detect.

Ten agent types are supported: keyword_detection, keyword_post_search, job_posting_keyword_detection, job_update, champion_tracking, plus five semantic watchlist types.

Note: For the five semantic watchlist types, a watchlist of the matching type is created automatically and bound to the new agent. Pass watchlist_id to bind an existing list instead — its type must equal the chosen agent type, otherwise the request returns 422.

champion_tracking works the same way, but always binds a profile watchlist of type champion — the same list a champion agent can bind to, simultaneously. Pass watchlist_id to bind an existing champion watchlist instead of creating one automatically; it must be of type champion, otherwise the request returns 422.

Competitor agents additionally accept an optional parameters.mode that decides what the agent does: lead_discovery (watches the competitor's team on LinkedIn and surfaces both directions — the people they comment on or react to, and the people who comment on or react to their own posts — always together, wherever those people work), account_monitoring (the default — only detects interactions between the competitor and your target accounts, in both directions, no scraping), or all (both). No other agent type accepts mode.

They also accept an optional parameters.target_group, independent of mode: it picks which people at the tracked competitor a lead-discovery run watches, and so whose LinkedIn interactions it reads. prospects watches their new-business side: sales, account executives, SDR/BDR, revenue and exec leadership; customers watches their account side: account management, customer success, support; all (the default when omitted) watches both.

competitor_activity and content_engagement are not among the ten types — they are labels that appear only on individual signal detections (see the Signals endpoint), never as an agent's own type, and cannot be passed here. A competitor agent running lead discovery reports results under both labels in the same run — one per direction — while its own type stays competitor.

Minimum setup required

Most agent types need a usable persona and at least one target account before they can start monitoring — otherwise there is nothing to target. Before creating an agent of one of these types, your workspace must have:

• a persona with both a job title and a location set (having a persona on file is not enough on its own), and
• at least one target account.

job_update, job_posting_keyword_detection, and champion_tracking are exempt and can be created right away, since none of them is tied to a specific persona or account list.

If your workspace does not meet these requirements yet, the request is refused with 422 and a message naming the missing prerequisite and the tool to call first, for example:

• "Create a persona with sillage_v2_upsert_persona (job_title and location required) before creating an agent."
• "Persona is missing job_title. Update it with sillage_v2_upsert_persona before creating an agent."
• "Add at least one top account with sillage_v2_add_top_accounts first."

The response also includes a type field that identifies the missing prerequisite, so your code can check it instead of parsing the message: https://docs.getsillage.com/errors/persona-incomplete when the persona is missing or incomplete (also returned when both prerequisites are missing — the message then names both), or https://docs.getsillage.com/errors/insufficient-top-accounts when only the target account is missing.

{
  "type": "https://docs.getsillage.com/errors/persona-incomplete",
  "title": "Persona incomplete",
  "status": 422,
  "detail": "Create a persona with sillage_v2_upsert_persona (job_title and location required) before creating an agent.",
  "instance": "/api/v2/agents"
}

Create agent

Request body

Example request — keyword detection

curl -X POST "https://api.getsillage.com/api/v2/agents" \
  -H "Authorization: Bearer $SILLAGE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "LinkedIn keyword monitor",
    "type": "keyword_detection",
    "parameters": {
      "tracking_keywords": ["fundraising", "series a", "\"series a\"", "hiring sales"]
    }
  }'

Example request — competitor watchlist (bound at creation)

curl -X POST "https://api.getsillage.com/api/v2/agents" \
  -H "Authorization: Bearer $SILLAGE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Competitor tracker",
    "type": "competitor",
    "watchlist_id": 7,
    "parameters": {
      "mode": "all",
      "target_group": "prospects"
    }
  }'

Example request — job update

curl -X POST "https://api.getsillage.com/api/v2/agents" \
  -H "Authorization: Bearer $SILLAGE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Job change tracker",
    "type": "job_update"
  }'

Example request — job posting keyword detection

curl -X POST "https://api.getsillage.com/api/v2/agents" \
  -H "Authorization: Bearer $SILLAGE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Hiring intent monitor",
    "type": "job_posting_keyword_detection",
    "parameters": {
      "tracking_keywords": ["salesforce", "revenue operations", "outbound"]
    }
  }'

Example request — champion tracking (implicit watchlist)

curl -X POST "https://api.getsillage.com/api/v2/agents" \
  -H "Authorization: Bearer $SILLAGE_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Champion job change tracker",
    "type": "champion_tracking"
  }'

Example response

201 Created

{
  "data": {
    "id": 42,
    "name": "LinkedIn keyword monitor",
    "type": "keyword_detection",
    "enabled": true,
    "schedulable": false,
    "schedule": null,
    "parameters": {
      "tracking_keywords": ["fundraising", "series a", "hiring sales"]
    },
    "watchlist_kind": null,
    "watchlist_id": null,
    "created_at": "2026-06-18T10:00:00.000Z",
    "updated_at": "2026-06-18T10:00:00.000Z"
  }
}

Response fields

Errors