Intelligence, trace and Solve commands

Read intelligence reports for sandboxes and fleets, list and fetch traces, send a local coding agent’s telemetry to Tangle Intelligence, and enter Solve competitions.

tangle intelligence sandbox my-box
tangle traces list
tangle connect claude-code --verify

tangle connect changes files on your machine. connect codex rewrites ~/.codex/config.toml without asking (pass --no-write to only print). connect claude-code offers to append a block to ~/.zshrc or ~/.bashrc, and writes ~/.zshrc when your shell is neither zsh nor bash. connect --verify waits in seconds, polling every 3 s, and exits 1 when nothing arrives before --timeout (default 120 s).

tangle traces reads the Intelligence API, so --base-url on it points at that host; TANGLE_INTELLIGENCE_BASE_URL sets it for every call.

Commands on this page

Command	What it does
`tangle intelligence connect`	Connect local agent CLIs to Tangle Intelligence
`tangle intelligence doctor`	Check whether your agent traces carry CONTENT or are metadata-only
`tangle intelligence hook`	Emit one local agent hook event to Tangle Intelligence
`tangle intelligence sandbox`	Create an intelligence report for one sandbox
`tangle intelligence fleet`	Create an intelligence report for a sandbox fleet
`tangle intelligence create`	Create a trace intelligence report
`tangle intelligence get`	Get an intelligence report
`tangle intelligence list`	List intelligence reports
`tangle traces list`	List trace summaries (one row per trace), newest first
`tangle traces get`	Show one trace’s spans. Streams NDJSON to stdout with `--ndjson`.
`tangle traces runs list`	List eval-runs, newest first
`tangle traces runs get`	Show a single eval-run
`tangle connect`	Connect a local coding agent’s telemetry to Tangle Intelligence, content on (for SDK integration snippets, see ‘tangle agent connect’)
`tangle opencode-exporter`	Export a running OpenCode’s content to Tangle Intelligence (standalone shim)
`tangle solve clone`	Clone the pinned competition repository
`tangle solve run`	Run the competition checker and record its evidence
`tangle solve submit`	Submit checker evidence through the device account and Tempo
`tangle solve account`	Show the local Solve signing account
`tangle solve device list`	List your devices and whether each may submit jobs on Tempo
`tangle solve device revoke`	Permanently revoke a device; the local device is revoked with its own key

Intelligence reports

`tangle intelligence connect`

Connect local agent CLIs to Tangle Intelligence.

Installs native hook files for local agent CLIs in a workspace so they emit traces to Tangle Intelligence, then sends a smoke trace.

tangle intelligence connect [agent] [options]

Aliases: connect-agent

Flag	Default	Description
`--endpoint <url>`	`https://intelligence.tangle.tools/v1/otlp`	OTLP endpoint base
`--workspace <path>`	`.`	Workspace to write hook files into
`--service-name <name>`		OTel service.name label
`--environment <name>`	`local`	OTel deployment.environment label
`--dry-run`		Print planned writes without changing files
`--hook-command <command>`	`tangle intelligence hook`	Command native hooks should run
`--no-smoke`		Skip the install smoke trace

tangle intelligence connect codex --workspace ~/repo --no-smoke

--dry-run prints the planned writes; --no-smoke skips the smoke trace.

`tangle intelligence doctor`

Check whether your agent traces carry CONTENT or are metadata-only.

tangle intelligence doctor [options]

Flag	Default	Description
`--endpoint <url>`	`https://intelligence.tangle.tools/v1/otlp`	OTLP endpoint base

`tangle intelligence hook`

Emit one local agent hook event to Tangle Intelligence.

Emits one local agent hook event as OTLP; native agent hooks call it.

tangle intelligence hook [options]

Flag	Default	Description
`--agent <agent>`		Agent name. Required.
`--event <event>`		Hook event name. Required.
`--payload <json>`		Hook payload JSON
`--endpoint <url>`		OTLP endpoint base
`--service-name <name>`		OTel service.name label
`--environment <name>`		OTel deployment.environment label
`--session-id <id>`		Session id override
`--dry-run`		Build the OTLP request without sending it
`--non-blocking`		Do not fail the parent agent process

Reads the hook payload from piped stdin (max 65536 bytes) when --payload is absent.

`tangle intelligence sandbox`

Create an intelligence report for one sandbox.

tangle intelligence sandbox <sandbox-id> [options]

Flag	Default	Description
`--mode <mode>`	`deterministic`	deterministic \| agentic
`--max-usd <amount>`		Maximum customer charge for agentic analysis
`--metadata <json>`		Metadata JSON object

Request: POST /v1/intelligence/reports

tangle intelligence sandbox my-box

`tangle intelligence fleet`

Create an intelligence report for a sandbox fleet.

tangle intelligence fleet <fleet-id> [options]

Flag	Default	Description
`--mode <mode>`	`deterministic`	deterministic \| agentic
`--max-usd <amount>`		Maximum customer charge for agentic analysis
`--metadata <json>`		Metadata JSON object

Request: POST /v1/intelligence/reports

tangle intelligence fleet fleet_1

`tangle intelligence create`

Create a trace intelligence report.

Creates an intelligence report for a sandbox or fleet subject.

tangle intelligence create [options]

Flag	Default	Description
`--subject-type <type>`		sandbox \| fleet. Required.
`--subject-id <id>`		Subject identifier. Required.
`--mode <mode>`	`deterministic`	deterministic \| agentic
`--max-usd <amount>`		Maximum customer charge for agentic analysis
`--metadata <json>`		Metadata JSON object

Request: POST /v1/intelligence/reports

tangle intelligence create --subject-type fleet --subject-id fleet_1

`tangle intelligence get`

Get an intelligence report.

Shows one intelligence report.

tangle intelligence get <job-id>

Request: GET /v1/intelligence/reports/{jobId}

tangle intelligence get job_1

`tangle intelligence list`

List intelligence reports.

Lists intelligence reports.

tangle intelligence list [options]

Flag	Default	Description
`--subject-type <type>`		sandbox \| fleet
`--subject-id <id>`		Subject identifier
`--limit <count>`		Maximum reports to return

Request: GET /v1/intelligence/reports

tangle intelligence list

Traces and runs

`tangle traces list`

List trace summaries (one row per trace), newest first.

Lists trace summaries, newest first, with filters.

tangle traces list [options]

Flag	Default	Description
`--from <iso>`		ISO-8601 lower bound on received time (inclusive)
`--to <iso>`		ISO-8601 upper bound on received time (inclusive)
`--model <model>`		Exact model match (any span carried this model)
`--run <runId>`		Exact run id match
`--status <status>`		ERROR \| OK
`-q, --query <text>`		Substring over span name
`--cursor <cursor>`		Opaque pagination cursor from a prior page
`--limit <count>`		Page size (clamped server-side to [1, 200])

Request: GET /v1/traces

tangle traces list

Global --base-url names the Intelligence API here; TANGLE_INTELLIGENCE_BASE_URL also sets it.

`tangle traces get`

Show one trace’s spans. Streams NDJSON to stdout with --ndjson.

Shows one trace’s spans; --ndjson streams all spans as NDJSON.

tangle traces get <traceId> [options]

Flag	Default	Description
`--ndjson`		Stream the full span set as NDJSON to stdout
`--cursor <cursor>`		Opaque pagination cursor from a prior page
`--limit <count>`		Spans per page

Request: GET /v1/traces/{traceId}/spans

tangle traces get trace_1 --json

--ndjson reads GET /v1/traces/{traceId}/spans.ndjson.

`tangle traces runs list`

List eval-runs, newest first.

Lists eval runs, newest first.

tangle traces runs list [options]

Flag	Default	Description
`--status <status>`		Run status filter
`--gate <decision>`		Promotion-gate decision filter
`--label <key:value>`		Match over the run’s labels
`--from <iso>`		ISO-8601 lower bound on received time
`--to <iso>`		ISO-8601 upper bound on received time
`-q, --query <text>`		Substring over run dir
`--cursor <cursor>`		Opaque pagination cursor from a prior page
`--limit <count>`		Page size

Request: GET /v1/runs

tangle traces runs list

`tangle traces runs get`

Show a single eval-run.

Shows one eval run.

tangle traces runs get <runId>

Request: GET /v1/runs/{runId}

tangle traces runs get run_1

Telemetry from a local coding agent

`tangle connect`

Connect a local coding agent’s telemetry to Tangle Intelligence, content on (for SDK integration snippets, see ‘tangle agent connect’).

Configures a local coding agent (claude-code, codex, opencode, openclaw, hermes, nanoclaw) to send telemetry with content to Tangle Intelligence; --verify checks that content arrives.

tangle connect [agent] [options]

Flag	Default	Description
`--endpoint <url>`	`https://intelligence.tangle.tools/v1/otlp`	OTLP endpoint base
`-p, --project <key>`		Project label (tangle.subject.key) to group runs; omit to set it yourself
`--rc <path>`		Shell rc file to append the Claude Code env block to
`--no-write`		Print the config instead of writing files (Claude Code / Codex)
`--inline-key`		Embed the resolved API key literally in the config (default: keep $`{TANGLE_API_KEY}`)
`--verify`		Poll onboarding status and report whether CONTENT is flowing
`--timeout <seconds>`	`120`	Verify poll timeout in seconds
`--run`		OpenCode only: run the content exporter against a live OpenCode server until interrupted
`--opencode-url <url>`		OpenCode server base URL for `--run` (or set OPENCODE_BASE_URL)
`--format <kind>`	`shell`	Claude Code output: shell (export block) \| env (.env / docker `--env-file`) \| dockerfile (ENV lines). env/dockerfile are the cron/container path, printed to stdout, no shell-rc edit, no TTY
`--content <tier>`	`full`	Claude Code content tier: full (everything) \| prompts (task+prompts+tool-args) \| metadata (telemetry on, content off)

This is Intelligence telemetry setup, not sandbox access; tangle agent connect prints SDK snippets.
codex: merges an [otel] block into ~/.codex/config.toml immediately, without a prompt, unless --no-write.
claude-code: prints an env block; on a terminal asks before appending it to the shell rc; --format env|dockerfile only prints.
--verify polls every 3 s up to --timeout SECONDS (default 120) and exits 1 unless content was seen.

`tangle opencode-exporter`

Export a running OpenCode’s content to Tangle Intelligence (standalone shim).

Runs until Ctrl-C, forwarding a running OpenCode server’s events with content to Tangle Intelligence.

tangle opencode-exporter [options]

Flag	Default	Description
`--endpoint <url>`	`https://intelligence.tangle.tools/v1/otlp`	OTLP endpoint base
`--opencode-url <url>`		OpenCode server base URL (or set OPENCODE_BASE_URL)

Same as tangle connect opencode --run.
OPENCODE_BASE_URL or --opencode-url names the OpenCode server.

Solve

`tangle solve clone`

Clone the pinned competition repository.

Clones a competition’s pinned repository.

tangle solve clone <competition> [directory] [options]

Flag	Default	Description
`--api-url <url>`		Solve API base URL
`--profile <name>`	`solve`	Tangle profile

`tangle solve run`

Run the competition checker and record its evidence.

Runs the competition checker in a directory and records its evidence.

tangle solve run [competition] [options]

Flag	Default	Description
`--directory <path>`		Competition directory (defaults to the current directory)

`tangle solve submit`

Submit checker evidence through the device account and Tempo.

tangle solve submit [competition] [options]

Flag	Default	Description
`--directory <path>`		Competition directory (defaults to the current directory)
`--pr-url <url>`		GitHub pull request containing the submission
`--model <name>`		Model used to produce the submission
`--note <text>`		Short note shown with the accepted submission
`--api-url <url>`		Solve API base URL
`--profile <name>`	`solve`	Tangle profile

`tangle solve account`

Show the local Solve signing account.

Shows the local Solve signing account; --create creates it.

tangle solve account [options]

Flag	Default	Description
`--create`		Create the account if it does not exist

tangle solve account --json

`tangle solve device list`

List your devices and whether each may submit jobs on Tempo.

tangle solve device list [options]

Flag	Default	Description
`--api-url <url>`		Solve API base URL
`--profile <name>`	`solve`	Tangle profile

`tangle solve device revoke`

Permanently revoke a device; the local device is revoked with its own key.

Permanently revokes a device after confirmation.

tangle solve device revoke <address> [options]

Flag	Default	Description
`--reason <text>`		Why the device is revoked (shown to the account)
`--yes`		Confirm the permanent revocation without a prompt
`--api-url <url>`		Solve API base URL
`--profile <name>`	`solve`	Tangle profile

Without a terminal it refuses unless --yes.

Hub and line commands Sandbox commands

Intelligence, trace and Solve commands

Commands on this page

Intelligence reports

tangle intelligence connect

tangle intelligence doctor

tangle intelligence hook

tangle intelligence sandbox

tangle intelligence fleet

tangle intelligence create

tangle intelligence get

tangle intelligence list

Traces and runs

tangle traces list

tangle traces get

tangle traces runs list

tangle traces runs get

Telemetry from a local coding agent

tangle connect

tangle opencode-exporter

Solve

tangle solve clone

tangle solve run

tangle solve submit

tangle solve account

tangle solve device list

tangle solve device revoke

Next

`tangle intelligence connect`

`tangle intelligence doctor`

`tangle intelligence hook`

`tangle intelligence sandbox`

`tangle intelligence fleet`

`tangle intelligence create`

`tangle intelligence get`

`tangle intelligence list`

`tangle traces list`

`tangle traces get`

`tangle traces runs list`

`tangle traces runs get`

`tangle connect`

`tangle opencode-exporter`

`tangle solve clone`

`tangle solve run`

`tangle solve submit`

`tangle solve account`

`tangle solve device list`

`tangle solve device revoke`