tapes

agentplay is an open-source command-line tool and local web viewer maintained by Entelligence AI for replaying coding-agent sessions as a map and timeline. It discovers session data from Claude Code, Codex, Cursor, GitHub Copilot, OpenCode, and Kiro, parses transcripts or local database stores into redacted agentplay/v1 JSON maps, and serves them from a local viewer. The session map records files and repositories touched, shell verbs, tools, MCP servers, models, subagents, errors, retries, delays, outputs, token counts, and timing without storing prompt or assistant text, tool inputs or outputs, file contents, or command arguments. The viewer provides a radial session map, playback with step or wall-clock time, error and retry navigation, task and subagent scopes, and import/export of map files. It runs locally without an account or API key; session data remains on the device, with an optional email-signup network request that can be disabled. The CLI requires Python 3.11 or newer, has no runtime dependencies beyond the standard library, and is distributed under the Apache License 2.0.

An agent skill for turning Claude Code sessions into a fine-tuned coding-agent model using Tinker and llama.cpp. Developed by Migel Tissera and distributed under the MIT license, it provides a nine-step workflow: collect and checksum transcripts, normalize and anonymize them, slice and compact sessions to the model context length, split datasets, train a LoRA on Tinker, compare checkpoints, merge the adapter, and convert and quantize the result to GGUF for local serving. The repository includes Python and shell scripts, reference notes, a synthetic transcript generator for offline dry runs, and support for Claude Code, OpenAI Codex, and OpenCode. It uses Tinker and tinker-cookbook for training and requires a Tinker API key for paid training; dataset preparation and smoke tests run locally.

agentacct is a local-first audit and observability tool for coding-agent work. It reads session logs from Claude Code, Codex, OpenCode, and Hermes, combines them with recorded tasks and test evidence, and produces a Work Receipt for each task covering commands run, files touched, tools used, time and token usage, cost, outcome, and evidence strength. Agent claims and machine verification are kept as separate axes, with verified status reserved for tasks whose live checks pass after the latest recorded work. The data is available through a macOS app, the live terminal dashboard launched with `agentacct tui`, and a local JSON API. State is stored in local files; the API listens only on loopback, and the project states that it has no login, cloud sync, or telemetry and does not request provider API keys.

telemetry.dev is an LLM observability platform for AI agents built on OpenTelemetry. It traces model calls, tool steps, retrievals, and MCP calls in a shared trace, exposing token usage, cost, latency, errors, and per-step timing. Applications can connect through TypeScript or Python SDKs or an existing OTLP/HTTP exporter. It provides filtering and search by status, environment, framework, or session, plus email alerts for costs, budgets, agent loops, and model or tool errors. Prompt capture can be disabled for individual environments.

Agent Console is an open-source, local-first observability tool by LockedIn Labs for AI coding agents. It reads Claude Code and Codex transcripts to show token usage by cache reads, cache writes, output and uncached input, model and session activity, list-price cost estimates, burn rate and dollars per hour. A self-hosted hub can combine privacy-limited usage metadata from multiple machines over TLS, while keeping prompts, replies, file paths and file contents local. It also provides optional Claude Code OpenTelemetry and Kong or LiteLLM metric ingestion, a Prometheus endpoint, Grafana dashboard, presenting mode, alerts, and policy hooks. It runs locally from source or release downloads on macOS, Linux and Windows, requires Node.js 22 or newer for the source route, and is licensed under the MIT License.