seclai-cli

Seclai CLI

Full-featured command-line interface for the Seclai platform. Manage agents, knowledge bases, sources, memory banks, evaluations, solutions, governance, and more — all from the terminal.

All commands return JSON to stdout, so you can pipe into jq or other tools.

Install

npm i -g @seclai/cli

Or run directly via npx (no install needed):

npx @seclai/cli agents list

Setup Skills and MCP

If you are using an AI coding agent like Claude Code, you can also install the skill individually with:

seclai skills install

This installs skills into all detected coding editors by default. Use --tool <tool> to scope it to one editor.

To install the Seclai MCP server into your editors (Cursor, Claude Code, VS Code, etc.):

seclai mcp configure --key "$SECLAI_API_KEY"

Or directly via npx:

npx skills add seclai/seclai-cli --full-depth --global --all
npx add-mcp https://api.seclai.com/mcp --header "X-API-Key: $SECLAI_API_KEY" --name Seclai

Documentation

Command reference (latest): https://seclai.github.io/seclai-cli/latest/

Authentication

The CLI supports two authentication methods:

API Key

Set the SECLAI_API_KEY environment variable, or pass --api-key per-command:

export SECLAI_API_KEY="sk-..."

# or inline
seclai --api-key "$SECLAI_API_KEY" agents list

SSO (OAuth2 Bearer Token)

SSO works out of the box with built-in production defaults — no configuration needed:

# Authenticate via browser (Authorization Code + PKCE)
seclai auth login

# Check authentication status
seclai auth status

# Refresh tokens manually
seclai auth refresh

# Log out (clears cached tokens)
seclai auth logout

For custom SSO settings (e.g. staging environment), use seclai configure sso or set environment variables:

Variable	Description	Default
SECLAI_SSO_DOMAIN	Cognito domain	auth.seclai.com
SECLAI_SSO_CLIENT_ID	Cognito app client ID	4bgf8v9qmc5puivbaqon9n5lmr
SECLAI_SSO_REGION	AWS region	us-west-2

Set profiles up interactively, and list the ones you have:

# Prompts for domain, client ID, region, and account ID
seclai configure sso [--profile-name <name>]

# Show every configured profile
seclai configure list

Use a named profile with --profile:

seclai --profile staging agents list

Tokens are cached in ~/.seclai/sso/cache/ and auto-refreshed when expired.

Environment Variables

Variable	Description
SECLAI_API_KEY	Default API key (alternative to --api-key)
SECLAI_API_URL	Override API base URL (default: https://api.seclai.com)
SECLAI_PROFILE	Default SSO profile name (default: default)
SECLAI_CONFIG_DIR	Config directory path (default: ~/.seclai)
SECLAI_API_VERSION	Dated API version (alternative to --api-version)
SECLAI_SSO_DOMAIN	Override SSO domain (default: auth.seclai.com)
SECLAI_SSO_CLIENT_ID	Override SSO client ID (default: 4bgf8v9qmc5puivbaqon9n5lmr)
SECLAI_SSO_REGION	Override SSO region (default: us-west-2)

Global Options

Flag	Description
--api-key <key>	Seclai API key
--profile <name>	SSO profile name
--account-id <id>	Account ID (X-Account-Id header)
--config-dir <path>	Config directory path
--api-version <date>	Opt into dated API changes released on or before this YYYY-MM-DD
--allow-unknown-api-version	Send an --api-version this CLI was not built against
--compact	Output compact (single-line) JSON
-V, --version	Print version

---

API Versions

The API is versioned by date. Responses can change shape between versions — a bare array becoming a {data, pagination} envelope, for instance — so the CLI sends no version header by default. Upgrading the CLI on its own never changes what a command prints.

# See which version a request resolves to
seclai api-version get

# Opt one invocation into the changes released up to a date
seclai --api-version 2026-07-27 alerts list

# Or for every client on the account, not just this CLI
seclai api-version set 2026-07-27
seclai api-version clear    # revert to the default

An --api-version this CLI was not built against is rejected, because a newer version can reshape a response the CLI would then misread. Pass --allow-unknown-api-version to send it anyway. api-version set is held to the same standard — it rejects a version this release was not built against, with the same override — because the pin applies to every client on the account and nothing re-checks it afterwards.

Each version includes the changes of the ones before it. seclai api-version --help prints the same list.

Version	What it changes
2026-07-01	The baseline, applied when no version is sent and the account is not pinned
2026-07-27	List responses move to {data, pagination}, and a query parameter the endpoint does not declare is rejected with a 422
2026-08-03	memory create and memory update reject a non-zero max_age_days, which reads as null; an omitted retention_days on create resolves per bank type
2026-08-21	sources create rejects an embedding dimension its embedder does not support — models embedders reports the supported ones
2026-09-28	Agent-definition writes such as agents def update use the current file-list grammar for a step's attachments
2026-09-30	A run's and a step's output, and a step's input, are the text rather than a JSON manifest. Files are in attachments on every version
2026-10-03	A new LLM step written without attachments takes its parent's files
2026-10-05	An item held for review by a governance policy reports content_status completed instead of failed; read held_for_review to find holds
2026-10-13	sources create and sources update reject a media_types value other than images or video, or any value on a source that is not a website or RSS feed, with a 400 where it was dropped
2026-10-16	An upload of any type is ready at once, and its extracted text is no longer written into a run's input
2026-10-18	A source's document_images defaults to true on a text-only embedder
2026-10-19	sources create rejects polling_max_items on any source but an RSS feed with a 400 where it was stored
2026-10-21	An evaluation result whose evaluator returned no verdict (status error) is left out of the failed counts, rates and average scores, and evals test-draft returns its score and passed as null instead of 0.5 and false

2026-09-30 changes what agents run and agents runs get print: a script that parses .output as JSON to find a run's files should read .attachments instead, which is present whichever version is in effect.

--api-key, --profile, --account-id and --config-dir reject an empty value. A shell expanding an unset variable passes "", which the SDK discards, so --api-key "$KEY" with KEY unset would fall back to SECLAI_API_KEY or a cached SSO session and run as a different identity — and --account-id "" would act on the default org rather than the one you named. Guard the flag instead of the value:

seclai ${KEY:+--api-key "$KEY"} agents list

An empty --api-version is still accepted, with a warning: it costs only the version header. A future release will reject that too.

---

Commands

Agents

seclai agents list [--page N] [--limit N]
seclai agents create --json '{"name":"My Agent"}'
seclai agents get <agentId>
seclai agents update <agentId> --json '{"name":"Renamed"}'
seclai agents delete <agentId>

# Pause an agent across every trigger path, then resume it
seclai agents disable <agentId>
seclai agents enable <agentId>

# Which live agents call this one via a call_agent step?
seclai agents callers <agentId>

Triggers

# Alias, sender allowlist, and inbound-handling flags for an EMAIL_RECEIVED trigger
seclai agents triggers email-config <agentId> <triggerId> --json '{"alias":"support"}'

# Named input attachments a dynamic_input or email_received trigger accepts.
# Steps reference them as {{agent.attachments.<name>}}; min defaults to 0.
seclai agents triggers accepted-attachments <agentId> <triggerId> \
  --json '{"accepted_attachments":[{"name":"invoice","content_types":["application/pdf"],"min":1,"max":1}]}'

Running Agents

Four modes: basic, streaming (SSE wait), events (NDJSON), and polling.

# Basic run (returns final result)
seclai agents run <agentId> --json '{"input":"Hello"}'
seclai agents run <agentId> --json-file ./run.json
cat run.json | seclai agents run <agentId> --json-file -

# SSE streaming (waits for done event or timeout)
seclai agents run <agentId> --json '{"input":"Hi"}' --stream --timeout-ms 60000

# NDJSON event stream (outputs every SSE event as a JSON line)
seclai agents run <agentId> --json '{"input":"Hi"}' --events
seclai agents run <agentId> --json '{"input":"Hi"}' --events --event-filter status
seclai agents run <agentId> --json '{"input":"Hi"}' --events --output data
seclai agents run <agentId> --json '{"input":"Hi","include_reasoning":true}' --events   # adds stream_reasoning events

# Polling (submit then poll until complete)
seclai agents run <agentId> --json '{"input":"Hi"}' --poll --poll-interval-ms 2000

Agent Runs

seclai agents runs list <agentId> [--page N] [--limit N]
seclai agents runs get <runId> [--include-step-outputs]
seclai agents runs cancel <runId>
seclai agents runs delete <runId>   # deprecated alias for `runs cancel`
seclai agents runs search [--page N] [--limit N] [--json '...']
seclai agents runs eval-results <agentId> <runId> [--page N] [--limit N] [--paged]
# Download a file attachment emitted by a run step. attachmentId is the
# URL-safe-base64 storage_key from run output manifests / webhooks.
seclai agents runs download-attachment <runId> <attachmentId> [--download-name <name>] [--output <path>]

Agent Definition

seclai agents def get <agentId>
seclai agents def update <agentId> --json '{"steps":[...]}'

Agent Export / Import

seclai agents export <agentId>
seclai agents export <agentId> --no-download

# Validate an agent_definition payload without creating an agent.
# Reports counts and any unresolved_refs you'll need to map with entity_remap
# when calling `agents create` or `agents update`.
# The body shape is `{ "agent_definition": <export payload> }`.
seclai agents preview-import --json-file ./preview-body.json
seclai agents export <agentId> \
  | jq '{agent_definition: .}' \
  | seclai agents preview-import --json-file -

Agent Input Upload

# Discover which files (if any) the agent expects before staging uploads.
seclai agents attachment-references <agentId>
seclai agents upload-input <agentId> --file ./data.csv [--file-name data.csv] [--mime-type text/csv]
seclai agents input-status <agentId> <uploadId>

Agent AI Assistant

seclai agents ai gen-steps <agentId> --user-input "Build a QA chatbot"
seclai agents ai step-config <agentId> --user-input "Configure the search step"
# --step-type is required: the API rejects the request without it.
seclai agents ai history <agentId> --step-type llm [--step-id <id>] [--limit N] [--offset N]
seclai agents ai mark <agentId> <conversationId> --json '{"accepted":true}'

Sources

seclai sources list [--page N] [--limit N] [--sort <field>] [--order asc|desc]
seclai sources create --json '{"name":"Docs","type":"manual"}'
seclai sources get <sourceId>
seclai sources update <sourceId> --json '{"name":"Renamed"}'
seclai sources delete <sourceId>

source is an alias for sources (e.g. seclai source list).

Upload

seclai sources upload <sourceId> --file ./doc.pdf [--title "Doc"] [--mime-type application/pdf] [--metadata '{}']
seclai sources upload-text <sourceId> --json '{"title":"Note","text":"Hello world"}'

Content Status

Whether what you uploaded has finished indexing. Both uploads return a content_version_id; pass it back to poll.

seclai sources contents list <sourceId> [--page N] [--limit N] [--sort created_at|title|status] [--order asc|desc]
seclai sources contents list <sourceId> --status failed
seclai sources contents list <sourceId> --content-version-id <id> --content-version-id <id>
seclai sources contents status <sourceId> <contentVersionId>

list prints {data, pagination} on every API version. Repeat --content-version-id to poll a batch in one request, and keep a request to about 100 ids: they travel in the URL, and one over 8,192 bytes is rejected with a 414. An empty id is refused rather than dropped, so an unset shell variable cannot widen the listing.

Email sources

# Mail an email source refused: sender and reason, never subject or body
seclai sources email-rejections <sourceId> [--page N] [--limit N]

To accept a refused sender, add its address or domain to the source's email_allowed_senders with sources update — which only helps when the rejection says authenticated is true.

Exports

seclai sources exports list <sourceId> [--page N] [--limit N]
seclai sources exports create <sourceId>
seclai sources exports get <sourceId> <exportId>
seclai sources exports cancel <sourceId> <exportId>
seclai sources exports delete <sourceId> <exportId>
seclai sources exports download <sourceId> <exportId>
seclai sources exports estimate <sourceId>

Embedding Migration

seclai sources migration get <sourceId>
seclai sources migration start <sourceId>
seclai sources migration cancel <sourceId>

Contents

seclai contents get <contentVersionId> [--start N] [--end N]
seclai contents delete <contentVersionId>
seclai contents embeddings <contentVersionId> [--page N] [--limit N]

Attachments

Files attached to a content item — email attachments, RSS enclosures from feed_attachments, web page media, the images inside a document — each with whether it is indexed and, when indexing failed, why.

seclai contents attachments list <contentVersionId> [--page N] [--limit N]
seclai contents attachments get <contentVersionId> <attachmentId> [--start N] [--end N]
seclai contents attachments embeddings <contentVersionId> <attachmentId> [--page N] [--limit N]
seclai contents attachments download <contentVersionId> <attachmentId> [--output <path>]

download answers 403 for a file withheld by a governance policy or still being indexed; get reports file_withheld first.

Upload / Replace

seclai contents upload <contentVersionId> --file ./updated.pdf [--metadata '{}']
seclai contents replace <contentVersionId> --file ./updated.pdf   # alias
seclai contents replace-text <contentVersionId> --json '{"title":"Note","text":"Updated content"}'

Knowledge Bases

seclai kb list [--page N] [--limit N] [--sort <field>] [--order asc|desc]
seclai kb create --json '{"name":"Support KB"}'
seclai kb get <kbId>
seclai kb update <kbId> --json '{"name":"Renamed"}'
seclai kb delete <kbId>

Memory Banks

seclai memory list [--page N] [--limit N]
seclai memory create --json '{"name":"Chat Memory","type":"conversation"}'
seclai memory get <memoryBankId>
seclai memory update <memoryBankId> --json '{"name":"Renamed"}'
seclai memory delete <memoryBankId>
seclai memory stats <memoryBankId>
seclai memory agents <memoryBankId>
seclai memory compact <memoryBankId>
seclai memory delete-source <memoryBankId>
seclai memory templates
seclai memory test-compaction <memoryBankId> [--json '...']
seclai memory test-compaction-standalone [--json '...']

Memory AI Assistant

seclai memory ai generate --user-input "Configure compaction"
seclai memory ai last
seclai memory ai accept <conversationId> --json '{"accepted":true}'

Evaluations

Criteria

seclai evals criteria list <agentId> [--page N] [--limit N] [--paged]
seclai evals criteria create <agentId> --json '{"name":"Quality"}'
seclai evals criteria get <criteriaId>
seclai evals criteria update <criteriaId> --json '{"name":"Renamed"}'
seclai evals criteria delete <criteriaId>
seclai evals criteria summary <criteriaId>

--paged wraps the results in {"data": [...]} instead of returning a bare array, so .data is a stable path to read whatever --api-version is in effect. The same flag is on alerts configs list, models alerts list and agents runs eval-results, whose top-level keys (configs, alerts) also become data from 2026-07-27. Nothing is invented: the pagination block appears only once the API sends one, from --api-version 2026-07-27. Migrate scripts to .data first, then opt in to get .pagination.

Results

seclai evals results list <criteriaId> [--page N] [--limit N]
seclai evals results create <criteriaId> --json '{"run_id":"...","score":0.9}'

Agent-level

seclai evals compatible-runs <criteriaId> [--page N] [--limit N]
seclai evals test-draft <agentId> --json '{"criteria":{...}}'
seclai evals agent-results <agentId> [--page N] [--limit N]
seclai evals agent-runs <agentId> [--page N] [--limit N]
seclai evals non-manual-summary <agentId>

Solutions

seclai solutions list [--page N] [--limit N]
seclai solutions create --json '{"name":"My Solution"}'
seclai solutions get <solutionId>
seclai solutions update <solutionId> --json '{"name":"Renamed"}'
seclai solutions delete <solutionId>

Link / Unlink Resources

seclai solutions link <solutionId> --agents '["id1","id2"]' --kb '["id3"]' --sources '["id4"]'
seclai solutions unlink <solutionId> --agents '["id1"]'

Conversations

seclai solutions convos list <solutionId>
seclai solutions convos add <solutionId> --json '{"user_input":"Add a source"}'
seclai solutions convos mark <solutionId> <conversationId> --json '{"accepted":true}'

Solution AI Assistant

seclai solutions ai generate <solutionId> --user-input "Add a FAQ source"
seclai solutions ai kb <solutionId> --user-input "Create a knowledge base"
seclai solutions ai source <solutionId> --user-input "Create a file source"
seclai solutions ai accept <solutionId> <conversationId>
seclai solutions ai decline <solutionId> <conversationId>

Governance

seclai governance ai generate --user-input "Create a content safety policy"
seclai governance ai list
seclai governance ai accept <conversationId>
seclai governance ai decline <conversationId>

Alerts

seclai alerts list [--page N] [--limit N] [--status open]
seclai alerts get <alertId>
seclai alerts status <alertId> --json '{"status":"resolved"}'
seclai alerts comment <alertId> --json '{"comment":"Investigating"}'
seclai alerts subscribe <alertId>
seclai alerts unsubscribe <alertId>

There is no --severity: GET /alerts declares no such filter, so it never filtered — it returned unfiltered rows that looked filtered. Filter client-side: seclai alerts list | jq '[.data[] | select(.severity == "high")]'.

Alert Configurations

seclai alerts configs list [--page N] [--limit N]
seclai alerts configs create --json '{"name":"Critical Alerts","type":"email"}'
seclai alerts configs get <configId>
seclai alerts configs update <configId> --json '{"name":"Renamed"}'
seclai alerts configs delete <configId>

Organization Alert Preferences

seclai alerts prefs list
seclai alerts prefs update <organizationId> <alertType> --json '{"enabled":true}'

Models

seclai models list [--provider <name>] [--supports-tool-use] [--supports-thinking]
seclai models list --supports-input-media image --supports-output-media video
seclai models get <modelId>

# Each media-generation modality and tier, with its model and cost
seclai models tiers

# Embedding models a source can index with, and rerankers a knowledge base can use
seclai models embedders [--supports-input-media text|image|video|audio] [--paged]
seclai models rerankers [--paged]

embedders and rerankers print the list under models, with the defaults and pricing beside it, on every API version. --paged prints it under data instead, plus the pagination block once the API sends one.

Model Alerts

seclai models alerts list [--page N] [--limit N]
seclai models alerts mark-read <alertId>
seclai models alerts mark-all-read
seclai models alerts unread-count

Recommendations

seclai models recommendations <modelId>

Model Playground Experiments

seclai models experiments list [--days N] [--start-date <date>] [--end-date <date>] [--limit N] [--offset N]
seclai models experiments create --json '{"model_ids":["gpt-4o"],"prompt":"Compare responses"}'
seclai models experiments get <experimentId>
seclai models experiments cancel <experimentId>
seclai models experiments delete <experimentId>   # soft-delete, preserves audit history

Search

seclai search --query "deployment guide" [--limit N] [--entity-type agent|source|kb]

# Search the Seclai documentation
seclai docs search --query "memory banks" [--mode keyword|semantic] [--limit N]

Account

# The authenticated user's account ID and organization memberships
seclai me

Email

Agent email: the domains agents send from, the inbound blocklist, inbound health, and recipient opt-outs.

Domains

seclai email domains list
seclai email domains add --kind custom --value mail.example.com [--delegated]
seclai email domains verify <domainId>        # check DNS now
seclai email domains set-primary <domainId>
seclai email domains test-email <domainId>    # send a test to the account owner
seclai email domains dmarc <domainId> [--days N] [--top-sources N]
seclai email domains remove <domainId>
seclai email domains use-shared               # revert to agent.seclai.com

add returns the DNS records to publish. Use --delegated when the domain's DNS is delegated to Seclai, so those records are published for you.

Blocked Senders

seclai email blocked list [--limit N] [--offset N]
seclai email blocked add --sender-email spam@example.com [--note "phishing"]
seclai email blocked add --sender-email example.com --match-type domain
seclai email blocked remove <blockedId>
seclai email blocked auto-block-mode disabled|input|input_and_output

Inbound Health

seclai email inbound status                   # quota, pause state, queued runs
seclai email inbound rejections [--agent-id <id>] [--limit N]
seclai email inbound cancel-queued            # fail all over-quota parked runs
seclai email inbound resume                   # lift the account-wide pause

Opt-outs

seclai email optouts list [--agent-id <id>] [--limit N] [--offset N]
seclai email optouts remove <optoutId>

Cloud Drives

The cloud-drive connections that file triggers, drive steps and cloud_drive sources read from. Connecting a drive happens in the app; these commands inspect and maintain the connections that exist.

seclai cloud-drives providers                 # providers that can be connected
seclai cloud-drives list
seclai cloud-drives get <connectionId>
seclai cloud-drives update <connectionId> [--name <name>] [--folder-path <path>]
seclai cloud-drives update <connectionId> --whole-drive
seclai cloud-drives agents <connectionId>     # agents using the connection
seclai cloud-drives rejections <connectionId> [--limit N]   # skipped files, newest first
seclai cloud-drives disconnect <connectionId> # revoke tokens, keep the connection
seclai cloud-drives delete <connectionId>

The listings print a plain array on every API version.

update changes only what you pass. Changing the folder resets the sync cursor, so files already in the new folder do not fire triggers. An empty --folder-path is refused, because the API reads it as the whole drive — say --whole-drive when that is what you mean.

delete is refused with a 409 while an agent trigger or a content source still depends on the connection. agents does not list content sources, so an empty result does not mean the delete will go through.

rejections is where to look when an agent did not run for a file: a skipped file fires no trigger. --limit takes 1 to 200 and defaults to 50.

AI Assistant

Top-level AI assistant for multi-domain operations.

seclai ai feedback --json '{"conversation_id":"...","feedback":"helpful"}'
seclai ai kb --user-input "Create a support knowledge base"
seclai ai source --user-input "Create a documentation source"
seclai ai solution --user-input "Build a customer support solution"
seclai ai memory --user-input "Create a conversation memory bank"
seclai ai memory-history
seclai ai accept <conversationId> [--json '...']
seclai ai decline <conversationId>
seclai ai memory-accept <conversationId> [--json '...']

Skills

Install Seclai skill files for AI coding tools (Copilot, Claude Code, Cursor, Windsurf, Codex, Kiro, Cline, Roo Code, Gemini, Antigravity).

# Auto-detect tools from workspace directory structure
seclai skills install

# Target a specific tool
seclai skills install --tool copilot
seclai skills install --tool claude
seclai skills install --tool cursor
seclai skills install --tool kiro
seclai skills install --tool cline

# Install for all supported tools
seclai skills install --tool all

# Specify a custom directory
seclai skills install --tool copilot --dir /path/to/project

# Via npx (no install required)
npx @seclai/cli skills install

Skills follow the Agent Skills specification. Each tool gets a seclai-cli/ directory containing a SKILL.md with YAML frontmatter and a references/ subdirectory for progressive disclosure:

Tool	Directory
Copilot	.github/copilot/seclai-cli/
Claude Code	.claude/skills/seclai-cli/
Cursor	.cursor/skills/seclai-cli/
Windsurf	.windsurf/skills/seclai-cli/
Codex	.codex/skills/seclai-cli/
Kiro	.kiro/steering/seclai-cli/
Cline	.clinerules/seclai-cli/
Roo Code	.roo/rules/seclai-cli/
Gemini	.gemini/seclai-cli/
Antigravity	.antigravity/seclai-cli/

You can also install skills using the skills CLI:

npx skills add seclai/seclai-cli

MCP Server

Configure the Seclai MCP server for AI coding tools:

# Auto-detect tools and write MCP config
seclai mcp configure --key YOUR_API_KEY

# Target a specific tool
seclai mcp configure --key YOUR_API_KEY --target claude-code
seclai mcp configure --key YOUR_API_KEY --target cursor
seclai mcp configure --key YOUR_API_KEY --target claude-desktop
seclai mcp configure --key YOUR_API_KEY --target windsurf

# Configure all known targets
seclai mcp configure --key YOUR_API_KEY --target all

# Show the MCP config snippet (for manual setup)
seclai mcp show
seclai mcp show --key YOUR_API_KEY

Target	Config File	Scope
claude-code	.mcp.json	Project
cursor	.cursor/mcp.json	Project
claude-desktop	~/Library/Application Support/Claude/claude_desktop_config.json	Global
windsurf	~/.codeium/windsurf/mcp_config.json	Global

The command merges into existing config files — it won't overwrite other MCP servers.

Shell Completion

Generate shell completion scripts for tab-completion of commands:

# Bash — add to ~/.bashrc
eval "$(seclai completion bash)"

# Zsh — add to ~/.zshrc
eval "$(seclai completion zsh)"

# Fish — save to completions directory
seclai completion fish > ~/.config/fish/completions/seclai.fish

---

Development

Install dependencies

npm install

Type checking

npm run typecheck

Build

npm run build

Run locally

npm run dev -- --help

Test

npm test

Test global install locally

npm run build
npm link
seclai --help