Developers

The production runtime for your AI agents.

Point any AI agent at AgentLed and get 100+ integrations, reusable code blocks, cache, retry, audit, and scoped permissions — without rewriting a line.

Works withClaude CodeCodex CLICursor AgentAny MCP client
Deploying agents for clients?See the developer partner program

What you get out of the box

Your agent handles the logic. AgentLed handles production.

One plan, every API

Stop juggling 10 API keys.

LinkedIn, Hunter, Specter, Affinity, Google Search, web scrapers — 100+ integrations on one subscription and one credit system. No more OAuth plumbing per workflow.

Compound your blocks

Scripts become permanent building blocks.

Save any step, sequence, or agent to your workspace. Reuse across workflows and share with teammates. Stop rewriting the same enrichment logic every session.

Production-ready

Cache, retry, audit, permissions.

Cache policies so expensive steps don't re-run. Retry-with-backoff on failure. Scoped per-integration permissions. Full input/output trace on every run.

The gap

What a production stack needs beyond the harness.

CapabilityCLI alone+ AgentLed
Integrations
Bring your own API keys and OAuth.
100+ pre-connected on one credit system.
Reusable steps
Ephemeral — scoped to the session.
Saved to workspace. Shareable with teammates.
Caching
Re-runs every step every time.
Cache policies per step.
Retries
Manual rerun on failure.
Automatic retry with backoff.
Permissions
All-or-nothing access.
Scoped, revocable per integration.
Observability
Terminal scrollback.
Structured audit trail per run, forever.
Scheduling
cron on your laptop.
Managed heartbeats. Runs 24/7.
Team sharing
Paste the script in Slack.
Workspace-scoped workflows and memory.

AgentLed CLI

The CLI lets you create, manage, and run workflows directly from your terminal. No browser needed.

Install

npx @agentled/cli setup

Create a workflow

agentled create "Outbound to fintech CTOs in Europe"

Run a workflow

agentled run outbound-eu --heartbeat 48h

Install in 30 seconds

Claude Code

claude mcp add agentled -e AGENTLED_API_KEY=wsk_... -- npx -y @agentled/mcp-server

OpenAI Codex

codex mcp add agentled -e AGENTLED_API_KEY=wsk_... -- npx -y @agentled/mcp-server
Get your API key atagentled.app→ Workspace Settings → Developer

What the MCP server exposes

100+ integrations on one credit

LinkedIn, Hunter, CRMs, web scraping, AI models, video generation — all callable from any MCP client. No per-service auth.

Persistent Knowledge Graph

Context persists across executions, sessions, and teammates. Your agent reads prior runs before making decisions, and writes results back for the next run.

Natural-language workflow build

Describe what you want in plain English. The server drafts a multi-step workflow, picks the right integrations, and saves it as a reusable block.

Enterprise-grade controls

GDPR compliant, EU data residency, no model training on your data, scoped permissions per integration, full audit trail.

New

MCP-051

Your AI agents now manage client relationships.

Every client gets a structured dossier — goals, contacts, meetings, decisions, open action items, linked workflows. Claude Code, Codex, and any MCP-aware agent read it before they act, write back what changed, and pick up where the last session ended.

Before the call

Brief the agent

Your agent reads the dossier and surfaces what's open: goals, risks, last decisions, pending follow-ups. No re-explaining the account every session.

During the meeting

Capture in structure, not in your head

New meeting notes drop in with sections the agent fills as you talk — summary, client feedback, decisions, action items, workflow changes, follow-up email draft.

After the meeting

Automated follow-up

The agent drafts the recap email, edits the workflows that changed, tracks action items until they close — across sessions, across teammates.

What your agent can do

agentled clients init acme --name "Acme Corp"

Open a new client dossier — goals, contacts, linked workflows, meeting log.

agentled clients brief acme

One-screen brief your agent reads before a meeting or hand-off.

agentled clients status acme

Open action items, latest meeting, decisions, linked workflows.

agentled clients meeting new acme --title "Q2 review"

Drop a structured note your agent fills in as the conversation happens.

agentled clients list

Every client your agent is currently managing, with last touch and open work.

Dossier layout

clients/
└── acme/
    ├── client.json        # goals, stage, next actions
    ├── contacts.json      # decision makers, roles
    ├── workflows.json     # linked workflow IDs
    ├── meetings/          # structured meeting notes
    ├── decisions/         # what was decided, when
    ├── feedback/          # raw client feedback
    └── replies/           # follow-up drafts

Why it matters

  • Plain JSON + Markdown — your agent edits dossiers like any other source file.
  • Lives in your workspace folder. No backend lock-in, no extra subscription.
  • Versions cleanly in git. Diff a quarter of client work.
  • Same primitive AgentLed engineers use when we build for clients.

Available tools

Workflows

list_workflowsget_workflowcreate_workflowupdate_workflowdelete_workflowvalidate_workflowpublish_workflow

Step Editing

get_stepadd_stepupdate_stepremove_stepmove_stepupdate_workflow_context

Import / Export

export_workflowimport_workflowpreview_n8n_importimport_n8n_workflow

Drafts & Snapshots

get_draftpromote_draftdiscard_draftlist_snapshotsrestore_snapshotcreate_snapshotdelete_snapshot

Executions

start_workflowlist_executionsget_executionlist_timelinesget_timelinestop_executionretry_executionrerun_step

Apps & Testing

list_appsget_app_actionstest_app_actiontest_ai_actionget_step_schema

Knowledge & Data

get_workspacelist_knowledge_listsget_knowledge_rowsget_knowledge_textquery_kg_edgesget_scoring_history

AI & Branding

chatget_brandingupdate_branding

Try it

Once connected, try this in Claude Code or Codex:

# List available integrations

"Show me what apps and actions I can use"

# Build a workflow

"Create a workflow that finds SaaS CTOs on LinkedIn, enriches emails via Hunter, scores by ICP, and saves to Knowledge Graph"

# Import from n8n

"Import my n8n workflow from this JSON and add AI scoring"

# Manage a client end-to-end

"Open the Acme dossier, summarise what's open, draft a follow-up email for last Tuesday's meeting, and add action items to track this week"

Skills & Patterns

agentic-ops

We've codified the workflow, scoring, and orchestration patterns that show up most consistently across customer builds as a Claude Code skill pack. Platform-agnostic, MIT-licensed, and contributed to by practitioners across the ecosystem — drop it into any Claude Code session to get the same patterns the AgentLed MCP server is steering toward.