AgentStack

Présentation

AgentStack is a hosted backend platform that an MCP-capable assistant can operate. Through one tool, agentstack.execute, it runs batched steps against projects covering data, auth, CRM, automation, knowledge, hosting, payments and scheduled tasks. Actions are found with discovery search and checked with a preflight step before writes.

It runs as a remote streamable HTTP endpoint at https://agentstack.tech/mcp. Work on your projects needs an AgentStack API key from Profile → API Keys, sent in the X-API-Key header (Authorization: Bearer with the same key also works). In Cursor, AgentStack's plugin can sign in with OAuth instead. To try it without an account, ask the assistant to create a project: AgentStack makes an anonymous one, limited to one project and 1,000 Logic Engine calls a month, and returns an API key to add to your client's headers. Each batch needs a project ID in its context. The server also ships prompts for logic rules, scheduled tasks, project setup and session setup.

Ce que vous pouvez faire

• Run batched AgentStack actions with one execute tool
• Discover available actions by describing a goal
• Run preflight checks before write operations
• Bind work to a specific project and manage its members
• Create backend logic rules and scheduled tasks
• Use built-in prompts for project and session setup

Outils 1

• agentstack.executePeut supprimer

  Run AgentStack operations as steps with action and params, within a project context; can write or…

Lus directement sur le serveur le 7 oct. 2026. Lecture seule, Écriture et Peut supprimer sont des indications déclarées par le serveur ; c’est votre client qui décide de vous demander confirmation avant d’exécuter un outil.

Exemples de prompts

• Create a new AgentStack project for my online course shop.

• Search AgentStack for actions that manage customers and list them.

• Add a scheduled task that runs a logic rule every morning.

• Add a teammate to my project as a read-only member.

Configuration

Chaque client se connecte à agentstack.tech. Choisissez le vôtre ; la page retient votre choix.

claude mcp add --transport http agentstack https://agentstack.tech/mcp --header "X-API-Key: <YOUR_API_KEY>"

{
  "mcpServers": {
    "agentstack": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://agentstack.tech/mcp",
        "--header",
        "X-API-Key: <YOUR_API_KEY>"
      ]
    }
  }
}

{
  "mcpServers": {
    "agentstack": {
      "url": "https://agentstack.tech/mcp",
      "headers": {
        "X-API-Key": "<YOUR_API_KEY>"
      }
    }
  }
}

{
  "servers": {
    "agentstack": {
      "type": "http",
      "url": "https://agentstack.tech/mcp",
      "headers": {
        "X-API-Key": "${input:x-api-key}"
      }
    }
  },
  "inputs": [
    {
      "type": "promptString",
      "id": "x-api-key",
      "description": "X-API-Key",
      "password": true
    }
  ]
}

{
  "mcpServers": {
    "agentstack": {
      "serverUrl": "https://agentstack.tech/mcp",
      "headers": {
        "X-API-Key": "<YOUR_API_KEY>"
      }
    }
  }
}

codex mcp add agentstack --url https://agentstack.tech/mcp

[mcp_servers.agentstack]
url = "https://agentstack.tech/mcp"
env_http_headers = { "X-API-Key" = "X_API_KEY" }

gemini mcp add --transport http --header "X-API-Key: <YOUR_API_KEY>" agentstack https://agentstack.tech/mcp

{
  "mcpServers": {
    "agentstack": {
      "type": "http",
      "url": "https://agentstack.tech/mcp",
      "headers": {
        "X-API-Key": "<YOUR_API_KEY>"
      }
    }
  }
}

Intégrer

Affichez AgentStack dans un README ou sur votre site. Chaque intégration renvoie à cette page, utilise l’URL du serveur et ne dépose aucun cookie. Comment fonctionnent les intégrations

La carte, l’iframe et le JSON sont en anglais. Ajoutez ?lang=it, fr, de ou es à leur adresse pour une autre langue.

Badge pour README Carte pour site web iframe JSON Prompt pour l’IA

Badge pour README

Markdown

[![AgentStack on mcp.tc](https://mcp.tc/i/agentstack/badge.svg)](https://mcp.tc/i/agentstack)

HTML

<a href="https://mcp.tc/i/agentstack"><img src="https://mcp.tc/i/agentstack/badge.svg" alt="AgentStack on mcp.tc" height="20"></a>

Il suit le mode clair ou sombre du lecteur. Ajoutez ?theme=light ou ?theme=dark pour le fixer, ou ?style=compact pour afficher le symbole sans le nom.

Carte pour site web

HTML

<script src="https://mcp.tc/w/agentstack.js" async></script>

La carte apparaît à l’endroit de la balise. Ajoutez data-theme="dark" ou data-size="compact" à la balise pour la modifier. Le fichier contient tout ce qu’il lui faut : il ne fait aucune autre requête et ne dépose aucun cookie.

iframe

HTML

<iframe src="https://mcp.tc/embed/agentstack" title="AgentStack on mcp.tc" width="420" height="200" loading="lazy" allow="clipboard-write" style="border:0;border-radius:8px;max-width:100%"></iframe>

Utilisez-la sur les pages qui n’autorisent pas les scripts. Ajoutez ?theme=light ou ?theme=dark à l’adresse pour fixer les couleurs.

JSON

La fiche sous forme de données pour vos pages et vos outils : nom, accroche, URL du serveur, outils et liens « Add to ». N’importe quel site peut la lire (CORS ouvert).

https://mcp.tc/i/agentstack.json

Ce que signifie chaque champ

Prompt pour l’IA

Collez-le dans un assistant de code IA, comme Claude Code ou Cursor, et il ajoutera la carte de AgentStack à votre site.

Markdown

Ajoute à mon site la carte mcp.tc de AgentStack, un serveur MCP. C’est un seul script statique : pas de build, aucun paquet à installer, aucun cookie.

Place cette balise là où la carte doit apparaître. La carte s’affiche juste avant elle :

```html
<script src="https://mcp.tc/w/agentstack.js" async></script>
```

Pour l’afficher ailleurs, mets ce conteneur à cet endroit et charge le même script une seule fois, par exemple avant `</body>`. Garde le lien à l’intérieur : les visiteurs sans JavaScript le voient.

```html
<div data-mcptc="agentstack"><a href="https://mcp.tc/i/agentstack">AgentStack on mcp.tc</a></div>
```

Options :
- `data-theme="light"` ou `"dark"` sur la balise ou le conteneur fixe les couleurs. Sans cela, la carte suit le mode clair ou sombre du visiteur.
- `data-size="compact"` donne une carte plus petite, sans l’accroche.
- La variable CSS `--mcptc-max-width` sur le conteneur ou un parent fixe la largeur (`520px` par défaut).
- `?lang=` dans l’URL du script choisit la langue de la carte : `en`, `it`, `fr`, `de` ou `es`.

Si le site a une Content Security Policy, autorise `script-src https://mcp.tc` et `img-src data:` (l’icône est intégrée).

Pour un README ou une autre doc en Markdown, utilise plutôt le badge :

```markdown
[![AgentStack on mcp.tc](https://mcp.tc/i/agentstack/badge.svg)](https://mcp.tc/i/agentstack)
```

Référence : https://mcp.tc/fr/embed

Historique des changements

Nous lisons la liste d’outils de ce serveur toutes les 3 heures et notons ce qui change. Dernière lecture : 8 oct. 2026, 07:29 UTC.

1. 7 oct. 2026, 22:33 UTC, Vérification automatique

   1 outil modifié.

   Détails, 7 oct. 2026, 22:33 UTC

   Modifiés 1

   • agentstack.executeschéma d’entrée et schéma de sortie
     • Description de context
       Avant

       aucun

       Après

       Execute scope. Set project_id on every batch after the session probe.

     • Description de options
       Avant

       aucun

       Après

       Batch controls: async jobs, recipes, playbook progress, and error policy.

2. 7 oct. 2026, 16:36 UTC, Vérification automatique

   1 outil modifié.

   Détails, 7 oct. 2026, 16:36 UTC

   Modifiés 1

   • agentstack.executedescription
     Avant

     AgentStack MCP = Work Graph control plane. After session + context.project_id: agents.work_next (mcp_work_loop_v1) -> plan_claim -> plan_execute (CAS/leases). Follow packet.next_action; catalog/discovery.search only if no packet. One-shot CRM/upload may call domain actions directly. Spine: agents.plan_get -> agents.work_next -> agents.plan_claim -> agents.plan_execute. Anti-pattern: Do not pick tools from catalog when packet.next_action is present. Mandatory session order: authenticate -> session probe auth.get_profile (alias auth.me) -> select/create project -> set context.project_id on every batch -> Work Graph loop. Playbook GET /mcp/prompts/get?name=agentstack_session_setup (recipe mcp_session_setup). Sync POST /mcp is lane mcp_batch (60000 ms) for the whole batch - not an RBAC cap. One heavy LLM step per sync execute (bots.simulate, knowledge.playground, agents.run with wait=true); that step widens to ai_stream (180000 ms). Large suites: options.async=true then poll discovery.job…

     Après

     AgentStack MCP = Work Graph control plane. After session + context.project_id: agents.work_next (mcp_work_loop_v1) -> plan_claim -> plan_execute (CAS/leases). Follow packet.next_action; catalog/discovery.search only if no packet. One-shot CRM/upload may call domain actions directly. Spine: agents.plan_get -> agents.work_next -> agents.plan_claim -> agents.plan_execute. Anti-pattern: Do not pick tools from catalog when packet.next_action is present. Mandatory session order: authenticate -> session probe auth.get_profile (alias auth.me) -> select/create project -> set context.project_id on every batch -> Work Graph loop. Playbook GET /mcp/prompts/get?name=agentstack_session_setup (recipe mcp_session_setup). Sync POST /mcp is lane mcp_batch (60000 ms) for the whole batch - not an RBAC cap. One expensive step per sync execute (bots.simulate, knowledge.playground, agents.run with wait=true, agents.plan_*, context.get); LLM-heavy steps widen to ai_stream (180000 ms). Large suites: options.a…

3. 7 oct. 2026, 01:33 UTC, Vérification automatique

   1 outil modifié. Instructions du serveur modifiées.

   Nouvelles instructions du serveur

   Détails, 7 oct. 2026, 01:33 UTC

   Modifiés 1

   • agentstack.executedescription
     Avant

     Bound project: agents.work_next (recipe mcp_work_loop_v1) before catalog search. discovery.search is fallback only. Mandatory session order: authenticate -> session probe auth.get_profile (alias auth.me) -> select/create project -> set context.project_id on every batch -> domain work. Playbook GET /mcp/prompts/get?name=agentstack_session_setup (recipe mcp_session_setup). Execute AgentStack operations as steps: id, action, params. Sync POST /mcp is lane mcp_batch (60000 ms) for the whole batch - not an RBAC cap. Heavy LLM actions (bots.simulate, knowledge.playground, agents.run with wait=true) : at most one per sync execute. A single heavy step is widened to ai_stream (180000 ms). Cheap discovery/list/get actions may batch (max 50 steps). Large suites: sequential executes, or options.async=true then poll discovery.job_status (REST GET /mcp/jobs/{{job_id}}). Catalog: GET /mcp/actions (see execute_budget). Cite one number: catalog_actions_public from GET /mcp/actions/summary. GET /mcp/he…

     Après

     AgentStack MCP = Work Graph control plane. After session + context.project_id: agents.work_next (mcp_work_loop_v1) -> plan_claim -> plan_execute (CAS/leases). Follow packet.next_action; catalog/discovery.search only if no packet. One-shot CRM/upload may call domain actions directly. Spine: agents.plan_get -> agents.work_next -> agents.plan_claim -> agents.plan_execute. Anti-pattern: Do not pick tools from catalog when packet.next_action is present. Mandatory session order: authenticate -> session probe auth.get_profile (alias auth.me) -> select/create project -> set context.project_id on every batch -> Work Graph loop. Playbook GET /mcp/prompts/get?name=agentstack_session_setup (recipe mcp_session_setup). Sync POST /mcp is lane mcp_batch (60000 ms) for the whole batch - not an RBAC cap. One heavy LLM step per sync execute (bots.simulate, knowledge.playground, agents.run with wait=true); that step widens to ai_stream (180000 ms). Large suites: options.async=true then poll discovery.job…

   Instructions du serveur

   Avant

   AgentStack MCP — one tool: agentstack.execute (POST /mcp). Bound project: agents.work_next (recipe mcp_work_loop_v1) before catalog search. discovery.search is fallback only. Authenticate first: GPT/Cursor OAuth Connect (user logs in), or POST /mcp/.well-known/oauth-token grant_type=password, or MCP auth.login. If this client cannot complete OAuth, use the user's API key: header X-API-Key or Authorization: Bearer <user API key> from the profile. Session probe: auth.get_profile (alias auth.me). Bind context.project_id before domain work. If session_ready or mutation_allowed, do not log in again and do not switch_project. Then agents.work_next. Call discovery.status only when the work packet has no next_action. discovery.search for the human goal, discovery.describe for one action, preflight.check before writes. Empty lists are valid. Diagnose before repair. A timeout retries the same idempotency key. Do not call a heavy_llm action only to read state. A page at /s/{id}/ uses the hosted cookie, not operator MCP auth.login. Playbook: {hints['prompts_url']}?name=agentstack_session_setup. When discovery.search returns response_mode=workflow, run that step. Do not treat catalog_page as the command. OAuth often binds ecosystem project_id=1 — override with tenant id.

   Après

   AgentStack MCP — one tool: agentstack.execute (POST /mcp). Bound project: agents.work_next (recipe mcp_work_loop_v1) before catalog search. discovery.search is fallback only. Authenticate first: GPT/Cursor OAuth Connect (user logs in), or POST /mcp/.well-known/oauth-token grant_type=password, or MCP auth.login. If this client cannot complete OAuth, use the user's API key: header X-API-Key or Authorization: Bearer <user API key> from the profile. Session probe: auth.get_profile (alias auth.me). Bind context.project_id, then agents.work_next. If session_ready or mutation_allowed, do not log in again and do not switch_project. discovery.search only when packet.next_action is absent. discovery.describe for one action. preflight.check before writes. Empty lists are valid. Diagnose before repair. A timeout retries the same idempotency key. Do not call a heavy_llm action only to read state. A page at /s/{id}/ uses the hosted cookie, not operator MCP auth.login. Playbook: /mcp/prompts/get?name=agentstack_session_setup. When discovery.search returns response_mode=workflow, run that step. Do not treat catalog_page as the command. OAuth often binds ecosystem project_id=1 — override with tenant id.

4. 6 oct. 2026, 22:22 UTC, Vérification automatique

   Début de l’observation : 1 outil.

FAQ