AI智能体与知识管理MCP服务

Seclai MCP Server 通过模型上下文协议将 AI 编程工具接入 Seclai,让用户能在编辑器或终端中直接管理智能体、知识库、内容源、记忆库、评估、治理、告警与模型。

  • by seclai
  • 混合部署
  • 1 星标
  • 更新于
在 GitHub 查看 seclai/seclai-mcp

使用说明

Seclai MCP Server

Connect AI-powered coding tools to Seclai using the Model Context Protocol. Manage your agents, knowledge bases, content sources, memory banks, evaluations, solutions, governance, alerts, models, and more directly from your editor or terminal.

Quick Start

Endpoint: https://mcp.seclai.com/mcp Transport: Streamable HTTP Auth: browser sign-in (OAuth), or an X-API-Key header

You need a Seclai account with API access.

Sign in from your client (recommended)

The server is an OAuth 2.1 authorization server with dynamic client registration and PKCE, so any client implementing the MCP authorization spec can register itself and open a browser sign-in. No API key to create, paste, or rotate, and no secret sitting in a config file.

Add the URL with no headers block at all and the client handles the rest. Most clients take this shape; each client's section below gives its exact format:

{
  "mcpServers": {
    "seclai": {
      "type": "streamable-http",
      "url": "https://mcp.seclai.com/mcp"
    }
  }
}

Claude Code, VS Code, Cursor and Codex CLI all support browser sign-in today. Claude (claude.ai and Desktop) adds the URL as a custom connector instead; see below. If a client reports an expired or invalid token, clear its stored credentials for the server and reconnect — it will re-register automatically.

What the server advertises
GET /.well-known/oauth-protected-resource/mcp
  resource:              https://mcp.seclai.com/mcp
  authorization_servers: https://mcp.seclai.com/

GET /.well-known/oauth-authorization-server
  authorization_endpoint:          /authorize
  token_endpoint:                  /token
  registration_endpoint:           /register     (dynamic client registration)
  grant_types_supported:           authorization_code, refresh_token
  code_challenge_methods_supported: S256
  scopes_supported:                openid

API key

For clients without OAuth support, or for CI and scripted use, create a key at seclai.com → Account Settings → API Keys and send it as X-API-Key. Take your client's entry from its section below and add a headers object to it:

"headers": {
  "X-API-Key": "YOUR_API_KEY"
}

Codex takes http_headers = { "X-API-Key" = "YOUR_API_KEY" } in its TOML entry, and Claude Code's CLI takes --header "X-API-Key: YOUR_API_KEY".

Keys must be user-scoped; keys created recently are user-scoped automatically.

To target a specific organization, include the X-Account-Id header.

On the URL. https://mcp.seclai.com/mcp is the canonical resource the server advertises in its protected-resource metadata. The bare host also routes, but prefer the /mcp form — OAuth clients derive the token audience from the resource identifier.

Claude (claude.ai and Desktop)

Add Seclai to Claude as a custom connector:

  1. Go to Customize → Connectors and click Add custom connector
  2. Enter the URL https://mcp.seclai.com/mcp and click Add
  3. Click Connect and sign in with your Seclai account

On Team and Enterprise plans, an organization Owner adds the connector under Organization settings → Connectors (Add → Custom), and each member then clicks Connect under Customize → Connectors to sign in with their own Seclai account.

Claude Code (CLI)

claude mcp add seclai \
  --transport http \
  "https://mcp.seclai.com/mcp"

Claude Code opens the browser sign-in on first use. Run /mcp inside Claude Code to check the connection or re-authenticate.

Or copy examples/claude-code.json to .mcp.json in your project root.

Cursor

Copy examples/cursor.json to .cursor/mcp.json in your project root.

Or add manually: Settings → MCP → Add new MCP server → set type to streamable-http and URL to https://mcp.seclai.com/mcp.

Windsurf

Copy examples/windsurf.json to ~/.codeium/windsurf/mcp_config.json.

VS Code (GitHub Copilot)

Copy examples/vscode.json to .vscode/mcp.json in your project, or run MCP: Open User Configuration for a user-level install. VS Code uses servers rather than mcpServers, and "type": "http":

{
  "servers": {
    "seclai": {
      "type": "http",
      "url": "https://mcp.seclai.com/mcp"
    }
  }
}

Kiro

Copy examples/kiro.json to .kiro/settings/mcp.json in your project, or ~/.kiro/settings/mcp.json for all projects.

Roo Code

Copy examples/roo.json to .roo/mcp.json in your project root. For a global install, use Roo's MCP Servers → Edit Global MCP panel.

Cline

Cline stores MCP settings in cline_mcp_settings.json inside its VS Code extension storage, so edit it through the UI rather than by path: open the Cline panel → MCP Servers → Configure MCP Servers, then add the block below (also in examples/cline.json). Cline names this transport streamableHttp; without a type it falls back to SSE:

{
  "mcpServers": {
    "seclai": {
      "type": "streamableHttp",
      "url": "https://mcp.seclai.com/mcp"
    }
  }
}

Gemini CLI

Add to .gemini/settings.json in your project, or ~/.gemini/settings.json globally (also in examples/gemini-cli.json). Gemini CLI names the field httpUrl, not url:

{
  "mcpServers": {
    "seclai": {
      "httpUrl": "https://mcp.seclai.com/mcp"
    }
  }
}

Codex CLI

Codex uses TOML at ~/.codex/config.toml, and can hold the OAuth credentials itself — see examples/codex.toml:

[mcp_servers.seclai]
url = "https://mcp.seclai.com/mcp"
auth = "oauth"

Then run codex mcp login seclai to complete sign-in. On older Codex builds that only discover stdio servers, add experimental_use_rmcp_client = true above the server block.

Antigravity

Copy examples/antigravity.json to ~/.gemini/config/mcp_config.json, or .agents/mcp_config.json in your workspace. Antigravity names the field serverUrl:

{
  "mcpServers": {
    "seclai": {
      "serverUrl": "https://mcp.seclai.com/mcp"
    }
  }
}

Other Clients

Any client supporting Streamable HTTP MCP transport works. Configure:

Setting Value
URL https://mcp.seclai.com/mcp
Transport Streamable HTTP
Auth (OAuth) none in config — the client registers and signs in
Auth (API key) X-API-Key: YOUR_API_KEY
Org targeting X-Account-Id: YOUR_ACCOUNT_ID (optional)

The URL field is not spelled the same everywhere — url in most clients, httpUrl in Gemini CLI, serverUrl in Antigravity — and VS Code uses servers where everyone else uses mcpServers.

Available Tools

Agents

Tool Description
list_agents List all agents in your account with pagination
get_agent Get an agent's metadata (name, description, trigger type)
get_agent_definition Get the full agent definition including steps and change_id
create_agent Create a new agent with an optional template
update_agent Update an agent's name and/or description
update_agent_definition Update an agent's definition with optimistic locking
delete_agent Delete an agent (cannot be undone)
export_agent Export an agent's portable JSON snapshot
preview_agent_import Validate an agent_definition payload without creating anything
disable_agent Pause an agent so it stops running from every trigger
enable_agent Re-enable a paused agent
get_agents_calling_agent List live agents that call this one via a call_agent step

Agent Runs

Tool Description
run_agent Start an agent run
get_agent_run_status Get the status and result of an agent run
list_agent_runs List runs for an agent with optional status filter
cancel_agent_run Cancel an in-flight or queued run by run id

Agent Inputs

Tool Description
get_agent_attachment_references What files an agent's definition expects on a run
upload_agent_input Upload a file to use as input for a subsequent run_agent
get_agent_input_upload_status Poll the processing status of an upload_agent_input
set_trigger_accepted_attachments Replace the files a trigger accepts (names, types, counts)

Call get_agent_attachment_references before staging uploads — it reports whether the agent accepts files and the names or patterns a run-time batch must satisfy.

Agent AI Assistant

Tool Description
generate_agent_steps Auto-generate agent steps from a natural-language description
generate_step_config Generate configuration for a specific step type
get_ai_conversation_history Get the AI assistant conversation history for an agent
mark_ai_suggestion Accept or decline an AI-generated suggestion

Evaluations

Tool Description
list_evaluation_criteria List evaluation criteria for an agent
get_evaluation_criteria Get details of a specific evaluation criteria
create_evaluation_criteria Create a new evaluation criteria
update_evaluation_criteria Update an existing evaluation criteria
delete_evaluation_criteria Delete an evaluation criteria
get_evaluation_summary Get summary statistics for an evaluation criteria
get_non_manual_evaluation_summary Account-wide summary of non-manual evaluations for a date range
list_evaluation_results List results for an evaluation criteria
list_agent_evaluation_results List all evaluation results for an agent
list_run_evaluation_results List evaluation results for a specific run
create_evaluation_result Create a manual evaluation result
run_evaluation Run automated evaluation on a criteria
test_draft_evaluation Test an evaluation criteria without persisting

Knowledge Bases

Tool Description
list_knowledge_bases List knowledge bases in your account with pagination
get_knowledge_base Get a knowledge base's details including its linked sources
create_knowledge_base Create a new knowledge base with at least one source
update_knowledge_base Update a knowledge base's name, description, or sources
delete_knowledge_base Delete a knowledge base (fails if referenced by agents)
generate_kb_plan AI-generate a knowledge base configuration plan

Knowledge Base Chat

Tool Description
chat_with_knowledge_base Chat with a knowledge base using natural language
list_kb_conversations List conversations for a knowledge base
load_kb_chat_messages Load messages for a knowledge base conversation
rename_kb_conversation Rename a knowledge base conversation
delete_kb_conversation Delete a knowledge base conversation

Content Sources

Tool Description
list_sources List all content sources in your account with pagination
get_source Get detailed information about a content source
create_source Create a new content source
update_source Update a source's name, polling interval, or retention
delete_source Delete a content source
get_source_content_text Get the text content of a source item
list_account_source_content List all source content in the account
list_source_pulls List pull (sync) operations for a source
get_source_pull_errors Get errors from a source pull operation
generate_source_plan AI-generate a source configuration plan
upload_source_content Upload an image/audio/video/PDF/text file to a custom index
list_source_content_status Indexing status of every item a source tried to index, including failures
list_content_attachments Files attached to a content item (email attachments, feed files, page media)
get_content_attachment One attached file: indexing status, metadata and indexed text
list_content_attachment_embeddings The chunks an attached file is indexed by
list_email_source_rejections Mail an email content source refused, and why

Source Exports

Tool Description
list_source_exports List exports for a source
get_source_export Get details of a source export
create_source_export Create a new source export
cancel_source_export Cancel a running source export
download_source_export Download a completed source export
estimate_source_export Estimate the size in bytes of a source export

Source Embedding Migrations

Tool Description
get_source_embedding_migration Get status of an embedding migration
start_source_embedding_migration Start a new embedding migration
cancel_source_embedding_migration Cancel a running embedding migration

Memory Banks

Tool Description
list_memory_banks List memory banks with pagination
get_memory_bank Get a memory bank's details
create_memory_bank Create a new memory bank (conversation or general)
update_memory_bank Update a memory bank's name or configuration
delete_memory_bank Delete a memory bank
get_memory_bank_stats Get usage statistics for a memory bank
get_agents_using_memory_bank List agents that reference a memory bank
compact_memory_bank_now Trigger compaction of a memory bank
delete_memory_bank_source Delete the source backing a memory bank
list_memory_bank_templates List available memory bank templates
test_compaction_prompt Test a compaction prompt without applying

Memory Bank AI

Tool Description
generate_memory_bank_config AI-generate a memory bank configuration
get_memory_bank_ai_conversation Your recent memory-bank AI assistant turns, across all banks
accept_memory_bank_suggestion Mark a memory-bank AI suggestion accepted or declined (applies nothing)

Solutions

Tool Description
list_solutions List solutions with pagination
get_solution Get a solution's details
create_solution Create a new solution
update_solution Update a solution's name or description
delete_solution Delete a solution
link_solution_resources Link agents, KBs, or sources to a solution
unlink_solution_resources Unlink resources from a solution
generate_solution_plan AI-generate a solution configuration plan
accept_solution_plan Accept an AI-generated solution plan
decline_solution_plan Decline an AI-generated solution plan

Governance

Tool Description
get_governance_settings Get current governance settings
update_governance_settings Enable/disable governance or update settings
list_governance_settings List all governance settings
list_governance_policies List all governance policies
get_governance_policy Get a specific governance policy
create_governance_policy Create a new governance policy
update_governance_policy Update a governance policy
list_governance_policy_documents Browse available sample policy documents
get_governance_policy_document Get a specific sample policy document
get_governance_policy_usage Get usage stats for a policy
list_governance_policy_changes List change history for a policy
list_governance_evaluations List governance evaluations (filter by verdict)
resolve_governance_evaluation Resolve a flagged governance evaluation
bulk_resolve_governance_evaluations Resolve multiple evaluations at once
test_draft_governance_policy Test a policy draft without persisting
test_governance_policy Test a policy and persist the result
opt_evaluation_into_review Move a policy-test evaluation into the review queue
set_policy_knowledge_bases Link knowledge bases to a policy for RAG evidence
list_policy_knowledge_bases List knowledge bases linked to a policy
get_circular_knowledge_bases Check for circular KB references before linking
get_governance_stats Get aggregate governance metrics
get_governance_usage Get governance usage details
get_governance_credit_estimates Estimate governance credit costs
list_governance_audit_trail List governance audit trail entries
list_governance_conversations List governance AI conversations
generate_governance_plan AI-generate a governance plan from natural language
accept_governance_plan Accept an AI-generated governance plan
decline_governance_plan Decline an AI-generated governance plan
delete_governance_policy Soft-delete a governance policy
review_governance_coverage Generate a structured policy-coverage review for a scope
list_governance_redactor_chunks Per-chunk audit records under a parent REDACT evaluation

Alerts

Tool Description
list_alerts List alerts, filterable by status, agent or source connection
get_alert Get details of a specific alert
change_alert_status Change an alert's status (e.g. resolve)
add_alert_comment Add a comment to an alert
list_alert_configs List alert configurations
get_alert_config Get a specific alert configuration
create_alert_config Create a new alert configuration
update_alert_config Update an alert configuration
delete_alert_config Delete an alert configuration
list_organization_alert_preferences List org-level alert preferences
update_organization_alert_preference Update an org-level alert preference

Models

Tool Description
list_models List models grouped by provider with optional filters
get_model Get full details for a specific model
list_embedding_models Embedding models a content source can be created with
list_reranker_models Reranker models a knowledge base can use
get_model_recommendations Get recommendations for a model
list_model_alerts List model-related alerts
dismiss_model_alert Dismiss a model alert
dismiss_all_model_alerts Dismiss all model alerts
list_generation_tiers Media-generation quality tiers per modality, with cost
list_managed_providers Seclai-managed OAuth providers configured on this deployment

Playground Experiments

Tool Description
list_experiments List model playground experiments
get_experiment Get a specific experiment
run_experiment Run a model playground experiment
cancel_experiment Cancel a running experiment

Account & Organization

Tool Description
get_account_credit_usage Get credit usage for the account
get_account_llm_calls_summary Get LLM call summary for the account
list_organizations List organizations the user belongs to
switch_organization Switch to a different organization
get_retention_settings How long the account keeps agent traces and governance evaluations
set_retention_settings Change the agent-trace retention window (schedules permanent deletion)

Resource Exports

Tool Description
list_resource_exports List resource exports
get_resource_export Get a specific resource export
create_resource_export Create a resource export
cancel_resource_export Cancel a resource export
download_resource_export Download a completed resource export

Transfers

Tool Description
list_transfer_targets List available transfer targets
preview_transfer Preview a resource transfer between accounts
execute_transfer Execute a resource transfer

Agent Email — Domains

Tool Description
list_email_domains The account's vanity and custom agent-email domains
add_email_domain Add and provision a domain; returns the DNS records to publish
verify_email_domain Re-check SES and DNS verification right now
set_primary_email_domain Make a verified domain the account's primary sending domain
use_shared_email_domain Revert to the shared agent.seclai.com domain
send_email_domain_test Send a test message from a verified domain to the account owner
get_email_domain_dmarc DMARC aggregate-report summary for a domain
remove_email_domain Remove a domain and tear down its SES identity and DNS

Agent Email — Inbound

Tool Description
set_email_trigger_config Set an EMAIL_RECEIVED trigger's alias and sender allowlist
list_blocked_email_senders Senders blocked from triggering EMAIL_RECEIVED agents
block_email_sender Block a sender or domain (idempotent)
unblock_email_sender Remove a sender from the blocklist
set_email_auto_block_mode How governance BLOCK verdicts feed the blocklist
list_inbound_email_rejections Inbound mail discarded without running an agent, and why
get_inbound_email_status Whether inbound mail is paused, and the queued-run backlog
cancel_queued_email_runs Fail all over-quota parked inbound-email runs at once
resume_inbound_email Lift the account-wide inbound-email pause
list_agent_email_optouts Recipients who opted out of this account's agent email
remove_agent_email_optout Revoke an opt-out, letting the recipient receive mail again

MCP Clients

Seclai agents can call other MCP servers. These tools manage those saved connections — the Seclai server acting as an MCP client itself.

Tool Description
list_mcp_catalog Browse popular external MCP servers you can add
list_mcp_clients The account's saved external-MCP configurations
get_mcp_client Get a single MCP client by id
create_mcp_client Save a configuration for an external MCP server
update_mcp_client Update an existing MCP client
delete_mcp_client Soft-delete an MCP client
test_mcp_client_connection Connect and list the external server's tools
refresh_mcp_client_tools Re-list and persist the external server's tool catalog
describe_mcp_client_tools The cached tool catalog for one client, in compact form
disconnect_mcp_client Revoke a connected OAuth client's stored tokens, keeping it
list_mcp_client_calls Tool calls your agents made to this client during runs
get_agents_using_mcp_client Agents whose workflow references this MCP client

Human-in-the-Loop

Tool Description
list_pending_hitl_requests HITL requests in the account, paginated
get_hitl_request One request with its status, recipients and votes
vote_on_hitl_request Submit the calling user's vote on a pending request
cancel_hitl_request Cancel a pending request

Cloud Drives

Tool Description
list_drive_providers Providers whose OAuth app is configured on this deployment
list_cloud_drives The account's cloud-drive connections
get_agents_using_cloud_drive Agents whose workflow references this connection
list_cloud_drive_rejections Files a cloud-drive connection skipped, and why

Search

Tool Description
search Search across all resource types in the account
search_docs Search the Seclai product documentation by content

Resources

Resource URI Description
Agent Definition Schema seclai://schema/agent-definition Full schema reference for agent definitions
String Substitutions seclai://schema/string-substitutions All supported string substitution expressions

Usage Examples

Once connected, interact with Seclai through natural language:

"Show me all my Seclai agents"

"Create a new agent called 'Daily News Digest' with a schedule trigger and the summarizer template"

"Add a retrieval step before the display step in my Daily News Digest agent"

"Create a knowledge base called 'Podcast Archive' using my podcast sources"

"Show me all my content sources and change the news feed to poll hourly"

Agent Templates

When creating agents, specify a template for a pre-configured starting point:

Template Description
blank Empty agent with no steps
retrieval_example Demonstrates knowledge base retrieval
simple_qa Basic question-answering agent
summarizer Summarizes the input text
json_extractor Extracts structured data from content
content_change_notifier Monitors sources for changes and notifies
scheduled_report Generates reports on a schedule
webhook_pipeline Processes input and posts it to a webhook

Rate Limits

MCP calls are rate-limited per account based on your plan:

Plan Calls / Minute
Personal 25
Starter 30
Team 100
Pro 600

Exceeded calls return a retry_after value. Most AI assistants handle this automatically.

Troubleshooting

Error Fix
Rate limit exceeded Wait the indicated seconds, or upgrade your plan
Tools not appearing Restart client, and compare its entry with your client's section above
invalid_token For OAuth clients, clear stored credentials for the server and reconnect. For API keys, confirm the key is current and user-scoped
Conflict on definition update Retry — your assistant will re-read and re-apply

Links

License

MIT