complete` (LEGACY)

Status: DOCUMENTED + LEGACY + DEPRECATED + FAILED_VERIFICATION — the reference page still documents the endpoint as "[Legacy]", but on 2026-09-18 every live request (retired claude-2.1 and current claude-haiku-4-5-20251001) returned 400 invalid_request_error: "The /v1/complete endpoint has been deprecated. Please use the /v1/messages endpoint instead. See https://platform.claude.com/docs/en/api/messages for details." Treat it as effectively retired. Sources: https://platform.claude.com/docs/en/api/completions · https://platform.claude.com/docs/en/api/completions/create · https://platform.claude.com/docs/en/build-with-claude/working-with-messages (migration) · https://platform.claude.com/docs/en/api/versioning Machine-readable: generated/fragments/parameters/anthropic-complete-legacy.json, generated/fragments/endpoints/anthropic-core.json Last verified: 2026-09-18 (raws: tmp-live/anthropic-core/k1_complete_legacy.json, k2_complete_haiku.json)

# Documented request

text

POST /v1/complete   (x-api-key, anthropic-version: 2023-06-01, content-type: application/json, [anthropic-beta], [anthropic-workspace-id])

Field	Type	Req.	Notes
`model`	string	yes	reference enum lists current ids, but no current model serves this endpoint
`prompt`	string (min 1)	yes	must alternate `\n\nHuman: …\n\nAssistant:` turns
`max_tokens_to_sample`	integer ≥1	yes	equivalent of `max_tokens`
`stop_sequences`	string[]	no	model always stops on `\n\nHuman:`
`stream`	boolean	no	SSE `completion` events (incremental since 2023-06-01)
`temperature` / `top_p` / `top_k`	number	no	same deprecated semantics as Messages
`metadata.user_id`	string ≤512	no

# Documented response

json

{"id": "compl_018CKm6gsux7P8yMcwZbeCPw", "type": "completion", "completion": " Hello! My name is Claude.", "model": "claude-2.1", "stop_reason": "stop_sequence"}

stop_reason: stop_sequence (yours or the built-in \n\nHuman:) | max_tokens. No usage object.

# Live result (2026-09-18)

json

{"type":"error","error":{"type":"invalid_request_error","message":"The /v1/complete endpoint has been deprecated. Please use the /v1/messages endpoint instead. See https://platform.claude.com/docs/en/api/messages for details."},"request_id":"req_011CfBxeT7j1hh51N3Lv1Phn"}

HTTP 400 for both claude-2.1 and claude-haiku-4-5-20251001; x-should-retry: false.

# SDK support

SDK	State
Python `anthropic` 1.7.0	`client.completions` removed (AttributeError) — v1 migration dropped it
TypeScript `@anthropic-ai/sdk` 0.126.0	`client.completions.create({model, prompt, max_tokens_to_sample, stream?})` still typed; live → `BadRequestError` 400
CLI `ant`	no `completions` resource documented

# Migration to Messages

Text Completions	Messages
`prompt: "\n\nHuman: X\n\nAssistant:"`	`messages: [{"role": "user", "content": "X"}]`
system text before the first `Human:`	top-level `system`
`max_tokens_to_sample`	`max_tokens`
`"\n\nAssistant: prefix"` prefill	trailing `assistant` message (rejected on Claude 4.6+ → use `output_config.format` / instructions)
`completion` string	`content[0].text`
`stop_reason: stop_sequence` (built-in)	`end_turn`; custom → `stop_sequence` + `stop_sequence` field
SSE `completion` events	`message_start` / `content_block_delta` … (`docs/anthropic/streaming.md`)

Examples: examples/anthropic/messages/legacy_complete.{sh,py,ts} (all demonstrate the 400).