Every run, step by step.
Each turn folds into one line that says how long the agent worked and how it ended, and opens to every step. Trace lays the run out on a timeline, Agents shows who started whom, and Details holds its context, connectors, and changes, with Create PR one click away.
Version the cache keysatlas
Worked14mActive time
Estimated cost$0.62From tokens and list prices
Tokens1.2MInput and output
Calls9938 model, 61 tool, 2 failed
6 turns, 14m workedTurn 1 of 6, 48s, $0.04: Version the cache keysTurn 3 of 6, 4m 12s, 1 failed step, $0.21: Run the cache suiteTurn 5 of 6, 1m 40s, $0.08: Add the branch to the CI key
ModelToolsSubagentsFailedHeight is active time
StepsTurn 3 of 6Cost
Model response$0.03
Tool call
Subagent run
Tool callSlowest
Tool callFailed
Model response$0.03
Tool call
Model response$0.03
npm testTool call
Started9:42:16 AM
Duration2m 31s
Exit code0
Result42 passed
Timed by the runtime
Open in conversation
Each turn folds into one line.
A finished turn says how long the agent worked and how it ended. Open it for every step in plain words, with a failed call on a branch of its own and what changed at the end.
Turns and stepsPut the branch name in the CI cache key and run the cache tests.
Read 7 files about cache keys and CI
Edited src/cache/keys.ts
Edited .github/workflows/ci.yml
Ran 2 steps at once
Failed: npm run lint exited 1
Ran tests, 42 passed
Finished2 files changed42 tests passed
The CI key now starts with the branch name, so two branches no longer share cached rows. Lint flagged an unused import, which I removed before the tests ran.
See who started whom.
When a conversation starts helpers or messages another agent, Agents draws them as cards: what each one does, its state, and its cost. A helper that needs you stands out, and its card opens its conversation.
The Agents viewVersion the cache keys
Put the branch name in the CI cache key
$0.3114m
Explore
Reads, SearchFind every reader of the cache key
$0.063m
Test runner
ShellRun the cache suite on both branches
$0.11now
Codex in relay
Asked whether relay reads the old key
under $0.019m
StartedMessaged
Context and connectors, beside the conversation.
Show details opens a pane with the run's harness and models, how full its context is and how it grew across compactions, what its runtime recorded, and its connectors.
Work beside a conversationOverview
HarnessClaude Code
ModelConversation model
TokensAbout 1.2M input, 40k output
Messages86 messages
Context
62% used219.3k / 1M
System promptSystem toolsMCP toolsMemory filesMessages
Over this conversation10 requests, 1 compaction
Connectors
Hhansard12 callsConnected
Ggithub4 calls, 1 failedConnected
LlinearNeeds sign-in
Open a pull request from the run.
Create PR commits the conversation's files on a new branch without changing your checkout, pushes it, and opens the pull request with the GitHub CLI. It follows the checks and never merges.
Create a pull requestCreate a pull request
TitlePut the branch name in the CI cache key
Files to commitM.github/workflows/ci.ymlMsrc/cache/keys.ts
Committing 2 filesCommitted 2 files as 3f2a91c
Pushing to origin as fix/ci-cache-keyPushed to origin as fix/ci-cache-key
Opening the pull requestOpened the pull request
Put the branch name in the CI cache key#128Open
fix/ci-cache-key into main
Checks are runningAll checks passed
Open on GitHub
- Ways to read a run
- 4
- Readable, Source, Trace, and Agents when the run started any.
- Facts for every step
- 11
- From when it started and how long it took to its tokens, cost, and exit code.
- Steps to a pull request
- 3
- Commit, push, and open, each shown as it happens.
- Merge buttons
- 0
- Merging stays with you, on GitHub.
More of Hansard
Personal Sync
Keep your whole archive in storage you own, with every computer you use in step with the others.
Teams
Share whole conversations from the repositories you choose with your team, under rules every member can read.
Sharing
Send one conversation as a web page anyone can open, a PDF, Markdown, or a trajectory file.
Insights
Rank models on your own work with your own Pareto frontier, and see what you pay against list price and how the work turns out.
Running and Review
Run each project's agents from one place, with tabs, changes, and terminals, and answer everything that waits on you in one queue.
Graph
See how your projects, conversations, files, commits, and decisions connect, and the evidence behind each link.
Slack bot
Start Claude Code and Codex runs on your Mac from Slack, approve each one in its thread, and get digests where you choose.
AI assistants
Let Claude.ai, ChatGPT, and other assistants search your agent conversations and memories, with your approval on your Mac.
Phone companion
Follow your agents from your phone, answer what they ask, and revive or start them while they run on your computer.
Accounts
Keep every Claude and Codex account in one place with its limits, route each conversation to one with allowance left, and cap API key spend.
Quality of life
Keep working when an account reaches its limit, and give every agent the MCP servers and skills you installed in one.
Put it on the record.
