See every prompt, response, and tool call
Architect replays agent sessions as a scrollable timeline — assistant messages, file reads, greps, and shell commands with timestamps and duration.
Observe every coding-agent run in one place
Relay captures Cursor, Claude Code, and other agent sessions via CLI hooks — then surfaces timelines, tool calls, and org secrets in the Architect dashboard.
Free tier · 500 sessions/month · No credit card
cursor/acme-api#1842
- 0:00 Session started · Cursor · acme-api
- 0:04 read_file backend/src/handlers/auth.ts
- 0:12 grep requireAuthenticatedWorkspace
- 0:28 Adding workspace role check before deploy handler…
- 0:41 run_terminal_cmd npm run typecheck
- 0:52 Session completed · 14 events · 52s
500
Free sessions every month
3 min
Typical relay install setup
1
CLI to verify capture
7 days
Starter retention included
Sessions you can replay. Relay captures agent runs with tool-level detail — not screenshots or guesswork.
Why teams switch
Coding agents are powerful — and mostly invisible
What changes when agent runs are captured, searchable, and shared across your organization.
Without Relay
- Agent runs live on each laptop with no shared audit trail
- Tool calls and shell commands disappear when the session ends
- Security can't review what coding agents actually did
With Relay
- Every session streams to Relay with prompts, tools, and timing
- Architect dashboard shows one timeline across Cursor, Claude Code, and more
- Platform teams debug failures and prove compliance from one view
Why Relay
Agent observability
built for platform and security teams.
Capture coding-agent runs from developer machines, stream them to Relay, and review timelines in Architect — without standing up your own log pipeline.
One view across Cursor, Claude Code, and more
Relay normalizes hook events from different coding agents into a single session model so platform teams aren't chasing logs in three IDEs.
Install once with relay install
The Relay CLI wires local hooks, writes nested auto-collect captures, then relay sync uploads them before you roll observability out to the whole org.
relay install
relay sync
Credentials in a vault, not in dotfiles
Adapter keys and API tokens live in your Relay organization vault — scoped to admins, encrypted at rest, and injected at capture time.
Find the run that broke prod
Filter sessions by tool name, repo, or developer. Review exactly which files an agent touched before a bad deploy or security incident.
Who it's for
Teams shipping with Cursor, Claude Code, and other agents need a shared record of what those agents did.
-
Platform engineering
Standardize how coding agents run across repos. See which tools developers use and where sessions fail before they block a release.
-
Security review
Audit agent tool calls — file reads, shell commands, network requests — with org-scoped retention and export for compliance.
-
Agent debugging
Replay the session that produced a bad diff. Filter by grep, read, or terminal events to find the step that went wrong.
-
Team leads & managers
Understand agent adoption without standing over shoulders. Session volume and duration metrics show where AI assistance actually lands.
-
Multi-repo organizations
One Architect workspace for every adapter. Compare Cursor vs Claude Code usage and consolidate secrets in the org vault.
How it works
From local hooks to org-wide observability
Three steps: install the CLI, verify a capture, then review agent runs in Architect.
-
Install Relay
Run relay install in your repo. The CLI registers Cursor hooks (and other adapters) to stream session events.
-
Sync a capture
After a Cursor or Claude Code task, run relay sync to upload nested files from ~/.relay/captures/<agent>/<session>.json.
-
Review in Architect
Open the Architect dashboard — browse session timelines, filter tool calls, and manage org secrets.
Start capturing
Building headless? Read the docs →
Guides
How teams roll out agent observability
Step-by-step workflows for installing Relay, verifying capture, and reviewing sessions in Architect.
Browse all guidesReview sessions and tool calls in Architect
Use the Architect dashboard to inspect agent timelines, tool inputs, and failures across Cursor and Claude Code.
Validate agent capture before you roll out Relay org-wide
Run relay run, trigger a Cursor task, and confirm events in Architect before team rollout.
Install Relay and capture your first Cursor session
Run relay install, trigger a Cursor task, and confirm hook events land in ~/.relay/captures before you sync to the API.
Recently added
What's new
Updated
- Cursor hook adapter streams agent runs to Relay
- Architect dashboard session timelines with tool-call detail
- Org secrets vault for adapter credentials
- relay install and relay run CLI for local setup
Capture your first agent session in minutes.
Run relay install, trigger a Cursor task, and open Architect to see the timeline
— free, no credit card.
Pricing
Simple, Transparent Pricing
No hidden fees. Upgrade, downgrade, or cancel anytime.
Starter
Start capturing $0 / forever
Solo devs & small teams trying capture
What's included
- 500 agent sessions / month
- 7-day session retention
- Cursor hook adapter
- Architect dashboard timelines
- CLI install & local verify
Popular
Team
Choose Team $29 / month
Platform teams standardizing on agents
What's included
- 5,000 agent sessions / month
- 30-day session retention
- Multi-agent adapters (Cursor, Claude Code, …)
- Org secrets vault
- Tool-call search & filters
- Team member roles
Enterprise
Contact sales Custom /
Security review, SSO, and custom retention
What's included
- Custom session volume
- Extended or unlimited retention
- SSO & audit exports
- Dedicated support & SLA
- VPC / private deployment (roadmap) Soon
Compare plans
| Feature | Starter | Team | Enterprise |
|---|---|---|---|
| Agent sessions / month | 500 | 5,000 | Custom |
| Session retention | 7 days | 30 days | Custom |
| Architect dashboard | |||
| CLI (relay install / relay run) | |||
| Cursor adapter | |||
| Claude Code & Windsurf adapters | |||
| Org secrets vault | |||
| Tool-call timeline search | |||
| Team roles | |||
| SSO & audit exports | |||
| Dedicated support |
One session equals one captured agent run — from hook install through Architect review.
Enterprise
Need SSO, audit exports,
or custom retention?
Same capture pipeline and Architect dashboard — plus contracts, security reviews, and hands-on rollout for larger orgs. Most teams start on the free tier.
- Single sign-on SAML/OIDC for enterprise workspaces
- Custom retention Keep agent sessions as long as compliance requires
- Dedicated support Named contact for rollout and security review
FAQ
Questions people ask before signing up
What does Relay capture from coding agents?
Relay records agent sessions from tools like Cursor and Claude Code — prompts, assistant responses, tool calls (read, grep, shell), hook events, and timing. You get a searchable timeline in the Architect dashboard, not a black-box log file on each laptop.
How do I install Relay locally?
Run relay install in your repo. It wires Cursor hooks (and other adapters as they ship) to write nested auto-collect captures. Use relay sync to upload them before your team rolls it out org-wide.
Where are org secrets stored?
API keys and adapter credentials live in your Relay organization vault — encrypted at rest, scoped to workspace admins, and never checked into git. Developers authenticate with Relay CLI; secrets are injected at capture time, not pasted into agent config.
How long is session data retained?
Retention depends on your plan: Starter keeps 7 days, Team keeps 30 days, and Enterprise offers custom retention. You can export session timelines before they age out.
Does Relay work across multiple agent tools?
Yes. Relay normalizes events from Cursor, Claude Code, Windsurf, and other adapters into one session model so platform and security teams see a single audit trail — not siloed logs per IDE.
Is my session data kept private?
Sessions are isolated to your organization. Relay does not train models on your capture data or share timelines across tenants.
Ready to see what your agents actually do?
Install Relay, capture your first session, and open Architect — free tier includes 500 sessions every month.