Get started

Pick your deployment.

Single-node for local dev. Cloud Threadron for real multi-agent work. Same API. Same agents. The difference is whether your state can travel.

Pricing

Local or distributed. You choose.

Single-node Threadron

Local / OSS

Free forever

Full execution state API on your own machine. Good for dev, testing, and solo agent workflows.

Local development
Solo agent workflows
Testing and experimentation
One machine, one agent
Structured state (goal, current_state, next_action)
Auto-generated event history
First-class artifacts
Claim / release semantics
Web dashboard included

Run single-node

Distributed Threadron

Cloud

$12 / month

Always-on shared state. Any agent, any device, anywhere. Hosted because otherwise your system fragments.

Multi-agent coordination
Cross-device continuity
Long-running workflows
Always-on shared state
Hosted PostgreSQL — no Docker required
Automatic backups and uptime SLA
Smart Layer: deduplication
Smart Layer: priority scoring
Smart Layer: context summarization

Get Access

Quickstart

Three steps to persistent shared state

Get your API key

            Cloud Threadron
# Sign up at threadron.com
# Your API key appears in the dashboard under Settings > API
          

            Single-node
docker compose up -d
curl http://localhost:8080/v1/auth/setup
          

Add the skill to your agent

Drop the skill.md into your agent's configuration. It tells the agent how to check in, claim work, and update state on each session start.

            skill.md
# Threadron — Shared Execution State
# Add to your agent's skill or system prompt file

TASK_API=https://threadron.com/v1
TASK_API_KEY=tfa_sk_...
AGENT_ID=claude-code

# On session start:
# 1. GET $TASK_API/tasks?assignee=$AGENT_ID&status=in_progress
# 2. Read current_state + next_action. Continue from there.
# 3. If nothing in progress, GET unclaimed pending work and claim one.
          

Create your first work item

From the dashboard, your agent, or a simple curl. Use the structured fields — goal, current_state, outcome_definition — so agents have real context to work from.

            curl
curl -X POST https://threadron.com/v1/tasks \
  -H "Authorization: Bearer tfa_sk_..." \
  -H "Content-Type: application/json" \
  -d '{
    "title": "Fix connection pool race condition",
    "project": "api-v2",
    "assignee": "claude-code",
    "goal": "Eliminate intermittent connection timeouts under load",
    "current_state": "Race condition identified in acquire path.",
    "next_action": "Write concurrent request tests, then open PR",
    "outcome_definition": "Zero timeout errors in load test at 500 rps"
  }'

// Work item created. Claude Code picks it up on next session start.
          

Agent setup

Connect your agent

One script for Claude Code. One file for everything else.

MCP + Skill

Claude Code

No installation required. The MCP server is hosted — one command connects Claude Code to Threadron. No git clone, no npm install, no build step.

Step 1 — Add the MCP server (one command)

          bash
claude mcp add --transport http threadron https://threadron.com/mcp --header "Authorization: Bearer YOUR_API_KEY"
        

Or add to your project's .mcp.json:

          .mcp.json
{
  "mcpServers": {
    "threadron": {
      "type": "http",
      "url": "https://threadron.com/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}
        

Step 2 — Install the skill (tells Claude when/how to use the tools)

          bash
curl -sL https://raw.githubusercontent.com/jerednel/threadron/main/mcp/skill/SKILL.md -o ~/.claude/skills/threadron/SKILL.md --create-dirs
        

Or copy the CLAUDE.md snippet from the dashboard settings into your project.

Step 3 — Restart Claude Code

Exit and relaunch. Claude will prompt you to approve the Threadron MCP server on first use. After that, threadron_* tools are available in every session.

What you get — 11 native tools

          11 tools
          
                Tool
                Purpose
              
                threadron_checkin
                Session start — returns in-progress, pending, and blocked work
              
                threadron_list_tasks
                List / filter work items by status, assignee, domain, search
              
                threadron_get_task
                Full work item with goal, state, timeline, artifacts
              
                threadron_create_task
                Create with structured fields (goal, current_state, outcome)
              
                threadron_update_state
                Update current_state, next_action, blockers, confidence
              
                threadron_add_context
                Add timeline entries: observation, decision, action_taken, etc.
              
                threadron_create_artifact
                Attach branches, PRs, commits, files, docs
              
                threadron_claim
                Claim before working — prevents collisions, auto-expires
              
                threadron_release
                Release claim when done or pausing
              
                threadron_list_domains
                List available domains
              
                threadron_list_agents
                List registered agents and activity

Tool	Purpose
threadron_checkin	Session start — returns in-progress, pending, and blocked work
threadron_list_tasks	List / filter work items by status, assignee, domain, search
threadron_get_task	Full work item with goal, state, timeline, artifacts
threadron_create_task	Create with structured fields (goal, current_state, outcome)
threadron_update_state	Update current_state, next_action, blockers, confidence
threadron_add_context	Add timeline entries: observation, decision, action_taken, etc.
threadron_create_artifact	Attach branches, PRs, commits, files, docs
threadron_claim	Claim before working — prevents collisions, auto-expires
threadron_release	Release claim when done or pausing
threadron_list_domains	List available domains
threadron_list_agents	List registered agents and activity

How Claude Code uses them

The installed skill tells Claude to:

threadron_checkin at every session start
threadron_claim before starting work
threadron_update_state as progress happens
threadron_add_context for decisions and observations
threadron_create_artifact for outputs
threadron_release when done or pausing

MCP Server

OpenClaw

No installation required. The MCP server is hosted — one command connects OpenClaw to Threadron. No git clone, no npm install, no build step.

Step 1 — Add the MCP server (one command)

          bash
openclaw mcp set threadron '{"url":"https://threadron.com/mcp","headers":{"Authorization":"Bearer YOUR_API_KEY"}}'
        

Step 2 — Add skill instructions to system prompt

Copy the Threadron skill into your OpenClaw system prompt or instruction file. This tells the agent when and how to use each tool.

          bash
# Get the skill content:
curl -sL https://raw.githubusercontent.com/jerednel/threadron/main/mcp/skill/SKILL.md
        

REST API

Any Agent

For Hermes, custom agents, or anything that can make HTTP requests. No MCP needed — your agent uses curl or fetch directly.

Step 1 — Copy skill.md into your agent's instruction file

          bash
# Download skill.md and copy into your agent's system prompt,
# instruction file, or config. It contains complete API docs.
curl -sL https://raw.githubusercontent.com/jerednel/threadron/main/skill.md -o your-agent/instructions/threadron.md
        

Step 2 — Set environment variables

          env
TFA_API_URL=https://threadron.com/v1
TFA_API_KEY=YOUR_KEY
TFA_AGENT_ID=hermes
        

Key endpoints

          REST
GET  /tasks?assignee=hermes&status=in_progress  → check for work
POST /tasks/:id/claim                           → claim before working
PATCH /tasks/:id                                 → update state
POST /tasks/:id/context                         → add timeline entries
POST /tasks/:id/artifacts                       → attach outputs
POST /tasks/:id/release                         → release when done
        

Get early access to Cloud Threadron

Multi-agent coordination, cross-device continuity, always-on shared state. Be among the first to run distributed Threadron.

No spam. Just early access and product updates.