# AgentStack (agentstack.tech)

> Backend for AI agents: one batched tool over ~668 actions (hosting, storage, auth, payments).

- name: io.github.agentstacktech/agentstack
- version: 1.0.0
- connection class: R0 (autonomous)
- trust: 96/100
- quality: 84/100 (good)
- owner: not claimed
- descriptor: https://api.protogrid.dev/v1/servers/io.github.agentstacktech%2Fagentstack
- tools: https://api.protogrid.dev/v1/servers/io.github.agentstacktech%2Fagentstack/tools

**An agent can connect right now, no human step.** Remote endpoint, no authentication, reachable on the last probe.

## Connection (mcpServers)

```json
{
  "mcpServers": {
    "agentstack": {
      "type": "http",
      "url": "https://agentstack.tech/mcp"
    }
  }
}
```


## Trust

- hygiene: 100
- liveness: 98
- freshness: 100
- provenance: 90
- drivers: +repository +namespace-matches-repo +website +reachable +uptime-96% +updated-13d-ago
- Derived from observable signals (official registry feed and our own credential-free probes); no code audit performed.

## Quality 84/100 (good)

### protocol: 100 (weight 25)

- pass `protocol.modern`: supports 2026-07-28 (server/discover)
- pass `protocol.stateless`: answers without a session
- pass `protocol.transport`: streamable HTTP
- pass `protocol.list_ttl`: tool list is cacheable (ttlMs 3600000)

### authorization: n/a (weight 15)

- n/a `auth.prm`, `auth.as_metadata`, `auth.cimd`: no remote uses OAuth
- n/a `auth.secret_in_url`: no templated URL

### tool hygiene: 100 (weight 30)

- pass `tools.descriptions`: every tool has a description
- pass `tools.description_length`: descriptions are concise
- pass `tools.schemas`: every tool has an object input schema
- pass `tools.annotations`: 1 of 1 tools declare readOnlyHint or destructiveHint
- pass `tools.directory_hints`: every tool declares readOnlyHint, destructiveHint, idempotentHint and openWorldHint
- pass `tools.token_cost`: about 615 tokens to load every tool
- pass `tools.api_dump`: tools are not a one-to-one API dump

### stability: 25 (weight 15)

- warn `stability.changes`: 23 tool changes in 30 days
- fail `stability.rug_pull`: 1 tools kept their name but changed most of their description in 30 days; review before trusting them

### dependencies: n/a (weight 15)

- n/a `deps.known_vulns`, `deps.mcp_sdk_version`, `deps.resolvable`: no npm or PyPI package

- tools: 1, about 615 tokens to load them all
- tool set hash: 93e9a231469975a1a0693eb5f024f05d
- tools last changed: 2026-10-18T15:24:30.870Z
- badge: [![protogrid quality](https://protogrid.dev/badge/io.github.agentstacktech/agentstack.svg)](https://protogrid.dev/servers/io.github.agentstacktech/agentstack)
- Checks run on what our credential-free, read-only probes observe; tools are never called and no code audit is performed.

### Tool changes

- 2026-10-18 `agentstack.execute` changed (description, inputSchema)
  - before: 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. Context.project_id on every batch. Detail: prompts/get agentstack_session_setup and agentstack_execute_budget.
  - after: 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. context.project_id on every batch. Prompts: agentstack_session_setup, agentstack_execute_budget.
- 2026-10-17 `agentstack.execute` changed (inputSchema)
- 2026-10-16 `agentstack.execute` changed (description, inputSchema, outputSchema) **meaning changed**
  - before: 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. Cheap batch: work_next, plan_get, plan_claim. One/sync: plan_propose, plan_apply_proposal, plan_recovery_scan, context.get. Skip catalog when packet.next_action is set. Session: auth.get_profile -> project -> context.project_id on every batch. Setup GET /mcp/prompts/get?name=agentstack_session_setup. Packs: GET /mcp/actions execute_budget.packs (session.v1, batch.v1). Sync mcp_batch 60000ms whole batch; one heavy step per execute (agents.run wait=true; agents.plan_apply_proposal, agents.plan_propose, agents.plan_recovery_scan, context.get). LLM lane ai_stream 180000ms. async=true then discovery.job_status (GET /mcp/jobs/{job_id}). Single tool agentstack.execute; catalog_actions_public from /mcp/actions/summary. Prefer logic.* and integrations.*; not rules.*; not webhooks.*. Do not pick tools from catalog when packet.next_action is present.
  - after: 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. Context.project_id on every batch. Detail: prompts/get agentstack_session_setup and agentstack_execute_budget.
- 2026-10-16 `agentstack.execute` changed (inputSchema)
- 2026-10-15 `agentstack.execute` changed (title, description, inputSchema, annotations)
  - before: 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. Cheap and batchable on one sync execute: agents.work_next, agents.plan_get, agents.plan_claim. One per sync execute: agents.plan_propose, agents.plan_apply_proposal, agents.plan_recovery_scan, context.get. 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). Packs session.v1 and batch.v1: GET /mcp/actions field execute_budget.packs. Send context.project_id; other session fields default server-side. 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 and other heavy actions per execute_budget; agents.run with wait=true; agents.plan_apply_proposal, agents.plan_propose, agents.plan_recovery_scan, context.get). agents.work_next, agents.plan_get, and agents.plan_claim are cheap and may share that batch. LLM-heavy steps widen to ai_stream (180000 ms). Large suites: options.async=true then poll discovery.job_status (REST GET /mcp/jobs/{job_id}). Cite catalog_actions_public from GET /mcp/actions/summary. There is one execute tool (agentstack.execute); do not quote tools_count as 'N tools'. Do not invent rules.* or webhooks.* (prefer logic.* and integrations.*).
  - after: 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. Cheap batch: work_next, plan_get, plan_claim. One/sync: plan_propose, plan_apply_proposal, plan_recovery_scan, context.get. Skip catalog when packet.next_action is set. Session: auth.get_profile -> project -> context.project_id on every batch. Setup GET /mcp/prompts/get?name=agentstack_session_setup. Packs: GET /mcp/actions execute_budget.packs (session.v1, batch.v1). Sync mcp_batch 60000ms whole batch; one heavy step per execute (agents.run wait=true; agents.plan_apply_proposal, agents.plan_propose, agents.plan_recovery_scan, context.get). LLM lane ai_stream 180000ms. async=true then discovery.job_status (GET /mcp/jobs/{job_id}). Single tool agentstack.execute; catalog_actions_public from /mcp/actions/summary. Prefer logic.* and integrations.*; not rules.*; not webhooks.*. Do not pick tools from catalog when packet.next_action is present.
- 2026-10-14 `agentstack.execute` changed (outputSchema)
- 2026-10-14 `agentstack.execute` changed (outputSchema)
- 2026-10-14 `agentstack.execute` changed (inputSchema, outputSchema)
- 2026-10-13 `agentstack.execute` added
  - 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. Cheap and batchable on one sync execute: agents.work_next, agents.plan_get, agents.plan_claim. One per sync execute: agents.plan_propose, agents.plan_apply_proposal, agents.plan_recovery_scan, context.get. 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). Packs session.v1 and batch.v1: GET /mcp/actions field execute_budget.packs. Send context.project_id; other session fields default server-side. 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 and other heavy actions per execute_budget; agents.run with wait=true; agents.plan_apply_proposal, agents.plan_propose, agents.plan_recovery_scan, context.get). agents.work_next, agents.plan_get, and agents.plan_claim are cheap and may share that batch. LLM-heavy steps widen to ai_stream (180000 ms). Large suites: options.async=true then poll discovery.job_status (REST GET /mcp/jobs/{job_id}). Cite catalog_actions_public from GET /mcp/actions/summary. There is one execute tool (agentstack.execute); do not quote tools_count as 'N tools'. Do not invent rules.* or webhooks.* (prefer logic.* and integrations.*).
- 2026-10-13 `agentstack.execute` removed
  - 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. Cheap and batchable on one sync execute: agents.work_next, agents.plan_get, agents.plan_claim. One per sync execute: agents.plan_propose, agents.plan_apply_proposal, agents.plan_recovery_scan, context.get. 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). Packs session.v1 and batch.v1: GET /mcp/actions field execute_budget.packs. Send context.project_id; other session fields default server-side. 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 and other heavy actions per execute_budget; agents.run with wait=true; agents.plan_apply_proposal, agents.plan_propose, agents.plan_recovery_scan, context.get). agents.work_next, agents.plan_get, and agents.plan_claim are cheap and may share that batch. LLM-heavy steps widen to ai_stream (180000 ms). Large suites: options.async=true then poll discovery.job_status (REST GET /mcp/jobs/{job_id}). Cite catalog_actions_public from GET /mcp/actions/summary. There is one execute tool (agentstack.execute); do not quote tools_count as 'N tools'. Do not invent rules.* or webhooks.* (prefer logic.* and integrations.*).

Full history: https://api.protogrid.dev/v1/servers/io.github.agentstacktech%2Fagentstack/changes

## Remotes

- https://agentstack.tech/mcp (streamable-http, auth none, reachable true, uptime30d 0.9607843)

## Packages

_none_

## Tools (1)

- `agentstack.execute`: 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. context.project_id on every batch. Prompts: agentstack_session_setup, agentstack_execute_budget.