Negotiation

Ask whether this agent can handle a task

Orchestrators use this to pick the best agent for a task before sending a real message/send. The agent returns a capability score (0–1) weighted across skill match, I/O compatibility, performance, load, and cost.

A score below the caller's min_score is interpreted as "decline" — the orchestrator should route the task elsewhere. This endpoint never actually performs the task; it's advisory.

post/agent/negotiation

Request body

task_summarystring required

One-sentence description of the work.

task_detailsstring

Longer context, audience, constraints.

input_mime_typesstring[]

output_mime_typesstring[]

max_latency_msinteger

Client's latency tolerance.

max_cost_amountstring

Client's cost tolerance (currency-agnostic; agreed per-agent).

required_toolsstring[]

forbidden_toolsstring[]

min_scorenumber

Decline threshold. Scores below this are treated as "no".

Response

Negotiation response

can_handleboolean required

scorenumber required

reasonsstring[]

Short bullet rationale for the score.

estimated_latency_msinteger

estimated_cost_amountstring

Changes

Changed in 3 of the 12 revisions of this API.51915

Apr 21, 20268a22dc1800443176See the full diff
- ▲
  the request property weights became required
  request-property-became-required
- ▲
  removed the required property accepted from the response with the 200 status
  response-required-property-removed
- ▲
  removed the required property confidence from the response with the 200 status
  response-required-property-removed
- ●
  the weights/cost request property's max was set to 1.00
  request-property-max-set
- ●
  the weights/io_compatibility request property's max was set to 1.00
  request-property-max-set
- ●
  the weights/load request property's max was set to 1.00
  request-property-max-set
- ●
  the weights/performance request property's max was set to 1.00
  request-property-max-set
- ●
  the weights/skill_match request property's max was set to 1.00
  request-property-max-set
- ●
  the weights/cost request property's min was set to 0.00
  request-property-min-set
- ●
  the weights/io_compatibility request property's min was set to 0.00
  request-property-min-set
- ●
  the weights/load request property's min was set to 0.00
  request-property-min-set
- ●
  the weights/performance request property's min was set to 0.00
  request-property-min-set
- ●
  the weights/skill_match request property's min was set to 0.00
  request-property-min-set
- ●
  removed the optional property latency_estimate_ms from the response with the 200 status
  response-optional-property-removed
- ●
  removed the optional property matched_capabilities from the response with the 200 status
  response-optional-property-removed
- ●
  removed the optional property matched_tags from the response with the 200 status
  response-optional-property-removed
- ●
  removed the optional property queue_depth from the response with the 200 status
  response-optional-property-removed
- ●
  removed the optional property rejection_reason from the response with the 200 status
  response-optional-property-removed
- ●
  removed the optional property skill_matches from the response with the 200 status
  response-optional-property-removed
- ●
  removed the optional property subscores from the response with the 200 status
  response-optional-property-removed
- ○
  api operation id assessTaskCapability removed and replaced with
  api-operation-id-removed
- ○
  removed the non-success response with the status 400
  response-non-success-status-removed
- ○
  added the optional property estimated_cost_amount to the response with the 200 status
  response-optional-property-added
- ○
  added the optional property estimated_latency_ms to the response with the 200 status
  response-optional-property-added
- ○
  added the optional property reasons to the response with the 200 status
  response-optional-property-added
- ○
  added the required property can_handle to the response with the 200 status
  response-required-property-added
This revision also has 21 changes that name no endpoint, such as unreferenced schemas being removed. See the revision's changelog
Feb 12, 2026970d68a5c55d228See the full diff
- ▲
  the max_cost_amount request property type/format changed from number/ to string/
  request-property-type-changed
- ▲
  removed the required property overall_score from the response with the 200 status
  response-required-property-removed
- ●
  removed the optional property reasoning from the response with the 200 status
  response-optional-property-removed
- ●
  removed the optional property scores from the response with the 200 status
  response-optional-property-removed
- ○
  added the optional property latency_estimate_ms to the response with the 200 status
  response-optional-property-added
- ○
  added the optional property matched_capabilities to the response with the 200 status
  response-optional-property-added
- ○
  added the optional property matched_tags to the response with the 200 status
  response-optional-property-added
- ○
  added the optional property queue_depth to the response with the 200 status
  response-optional-property-added
- ○
  added the optional property rejection_reason to the response with the 200 status
  response-optional-property-added
- ○
  added the optional property skill_matches to the response with the 200 status
  response-optional-property-added
- ○
  added the optional property subscores to the response with the 200 status
  response-optional-property-added
- ○
  added the required property score to the response with the 200 status
  response-required-property-added
Feb 11, 2026975e49e9b6751See the full diff
- ○
  endpoint added
  endpoint-added

Full history