Conversations

Read a conversation transcript

The harness’s own transcript (Claude Code or Codex JSONL), as the conversation’s computer wrote it, kept by Shardflux as the conversation runs. Reading it never reaches the computer: it works the same for a working, parked or finished conversation. Page with offset = the previous next_offset; size is what is stored now.

get/v1/conversations/{id}/transcript

Request

  • Base URL: https://api.shardflux.dev
  • URL: https://api.shardflux.dev/v1/conversations/{id}/transcript
  • Auth: HTTP bearer

Path parameters

idstring required

Query parameters

offsetinteger
limit_bytesinteger

About how many bytes to return (whole lines; 0 returns the size only).

session_idstring

Response

Default Response

conversation_idstring required
harness'claude' | 'codex' required
session_idstring nullable required

The harness session the transcript belongs to (a Claude Code session id or Codex thread id); null while nothing is stored.

filestring nullable required

The native file name: <session>.jsonl for Claude Code, rollout-<time>-<thread>.jsonl for Codex.

offsetinteger required

Byte offset of data in the transcript.

next_offsetinteger required

Byte offset after data: pass it as the next offset. A line boundary.

sizeinteger required

Bytes stored so far.

datastring required

Native JSONL lines (UTF-8), byte for byte as the harness wrote them.

updated_atstring date-time nullable required

Changes

No recorded changes to this operation across all 1 revision of this API.