AgentStack (agentstack.tech) is an MCP server published by io.github.agentstacktech, hosted at agentstack.tech. When last checked on 2026-11-21 it was working: it answers and lists its tools. It declares 2 tools, including agentstack.execute, agentstack.read.
Backend for AI agents: one batched tool over ~668 actions (hosting, storage, auth, payments).
Nothing else in the archive shares these values.
Inferred from the host name and headers: what a server runs on, not who runs it.
AUTH-FIRST: auth.get_profile before mutate. IDE: Connect; no auth.login in chat. Unauth: auth_choices; skip work_next. Mutate: OAuth/PAT (mcp:execute). Anon: bootstrap_ack+create_project_anonymous. Spine: plan_get->work_next->plan_claim->plan_execute. WG: work_next->plan_claim->plan_execute; follow packet.next_action. Prompts: session_setup, execute_budget.
| Parameter | Type | Description |
|---|---|---|
| steps * | array | Steps; one heavy LLM per sync execute. |
| context | object | session.v1; project_id required. |
| options | object | batch.v1 controls. |
None yet. Changes appear after a server's second daily check.
Squares are this server's values; dots are other servers sharing them. A shared address or publisher is a lead, not proof of a common operator.
{
"name": "agentstack.execute",
"title": "AgentStack execute",
"description": "AUTH-FIRST: auth.get_profile before mutate. IDE: Connect; no auth.login in chat. Unauth: auth_choices; skip work_next. Mutate: OAuth/PAT (mcp:execute). Anon: bootstrap_ack+create_project_anonymous. Spine: plan_get->work_next->plan_claim->plan_execute. WG: work_next->plan_claim->plan_execute; follow packet.next_action. Prompts: session_setup, execute_budget.",
"inputSchema": {
"type": "object",
"required": [
"steps"
],
"properties": {
"steps": {
"type": "array",
"items": {
"type": "object",
"required": [
"id",
"action"
],
"properties": {
"id": {
"type": "string",
"description": "Step id in batch."
},
"action": {
"type": "string",
"description": "Dotted action from /mcp/actions."
},
"params": {
"type": "object",
"description": "Action params."
}
},
"description": "One catalog action."
},
"minItems": 1,
"description": "Steps; one heavy LLM per sync execute."
},
"context": {
"type": "object",
"properties": {
"project_id": {
"type": "integer",
"description": "Tenant project id."
}
},
"description": "session.v1; project_id required."
},
"options": {
"type": "object",
"properties": {
"goal": {
"type": "string",
"description": "Binds the batch to a playbook. A direct call uses execution_intent direct and omits goal. Per-step query must match this text."
},
"async": {
"type": "boolean",
"description": "Queue; poll discovery.job_status."
},
"envelope": {
"enum": [
"compact",
"full"
],
"type": "string",
"description": "compact (default) | full — ADR MCP_EXECUTE_ENVELOPE_COMPACT_V2"
},
"recipe_id": {
"type": "string",
"description": "MCP recipe id."
},
"stopOnError": {
"type": "boolean",
"description": "Stop on first failure."
},
"focus_node_id": {
"type": "string",
"description": "Work graph node id from agents.work_next packet."
},
"continueOnError": {
"type": "boolean",
"description": "Continue after failure."
},
"idempotency_key": {
"type": "string",
"description": "Retry receipt key."
},
"semantic_strict": {
"type": "boolean",
"description": "Reject batch_semantic_anchor_conflict when anchor set."
},
"execution_intent": {
"enum": [
"direct",
"workflow",
"validate"
],
"type": "string",
"description": "Batch lane: direct, workflow, or validate."
},
"plan_execute_mode": {
"type": "boolean",
"description": "Bind plan_execute lane for graph mutations."
}
},
"description": "batch.v1 controls."
}
},
"description": "batch.v1 execute. Direct call: options.execution_intent=direct and no goal. {action, params} is accepted and becomes one step. goal binds a playbook."
},
"outputSchema": {
"type": "object",
"properties": {
"ok": {
"type": "boolean",
"description": "False on total failure."
},
"job_id": {
"type": "string",
"description": "Poll id."
},
"timings": {
"type": "object",
"description": "Batch phase ms (hydrate_ms, preflight_ms, …).",
"additionalProperties": {
"type": "number"
}
},
"agent_os": {
"type": "object",
"description": "Follow next_step."
},
"recovery": {
"type": "object",
"description": "Next action when blocked."
},
"trace_id": {
"type": "string",
"description": "Support trace."
},
"error_code": {
"type": "string",
"description": "Stable failure code."
},
"execution_mode": {
"type": "string",
"description": "async when queued."
},
"response_bytes": {
"type": "integer",
"description": "Approximate serialized JSON size of the batch root."
},
"partial_success": {
"type": "boolean",
"description": "Some steps failed."
}
},
"description": "result.v1 batch envelope."
},
"annotations": {
"readOnlyHint": false,
"openWorldHint": true,
"idempotentHint": false,
"destructiveHint": false
}
}Read-only batch (effect read/evaluate). Same {steps, context, options} body as agentstack.execute. Server rejects mutate steps.
| Parameter | Type | Description |
|---|---|---|
| steps * | array | Same items as agentstack.execute. |
| context | object | session.v1. |
| options | object | Same keys as agentstack.execute options. |
{
"name": "agentstack.read",
"title": "AgentStack read",
"description": "Read-only batch (effect read/evaluate). Same {steps, context, options} body as agentstack.execute. Server rejects mutate steps.",
"inputSchema": {
"type": "object",
"required": [
"steps"
],
"properties": {
"steps": {
"type": "array",
"items": {
"type": "object",
"required": [
"id",
"action"
],
"properties": {
"id": {
"type": "string",
"description": "Step id."
},
"action": {
"type": "string",
"description": "Dotted action."
},
"params": {
"type": "object",
"description": "Action params."
}
},
"description": "One catalog action.",
"additionalProperties": true
},
"minItems": 1,
"description": "Same items as agentstack.execute."
},
"context": {
"type": "object",
"properties": {
"project_id": {
"type": "integer",
"description": "Tenant project id."
}
},
"description": "session.v1.",
"additionalProperties": true
},
"options": {
"type": "object",
"description": "Same keys as agentstack.execute options.",
"additionalProperties": true
}
},
"description": "Same body as agentstack.execute. Writes rejected.",
"additionalProperties": true
},
"outputSchema": {
"type": "object",
"properties": {
"ok": {
"type": "boolean",
"description": "False on total failure."
},
"trace_id": {
"type": "string",
"description": "Support trace."
},
"error_code": {
"type": "string",
"description": "Stable failure code."
},
"partial_success": {
"type": "boolean",
"description": "Some steps failed."
}
},
"description": "Same result.v1 envelope as agentstack.execute."
},
"annotations": {
"readOnlyHint": true,
"openWorldHint": false,
"idempotentHint": true,
"destructiveHint": false
}
}AgentStack MCP — one tool: agentstack.execute (POST /mcp). After VPS backend restart, first agents.work_next may fail while MCP tools import — wait ~15s and retry once (not a plan graph bug). Bound project: agents.work_next (recipe mcp_work_loop_v1) before catalog search. discovery.search is fallback only. Authenticate first: GET /mcp/connect → /mcp/.well-known/oauth-authorize (PKCE), IDE Device Code POST /api/oauth2/device/authorize, 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. Packs session.v1/batch.v1: discovery.list exposes execute_budget.packs. Work Graph skill: /mcp/prompts/get?name=agentstack_work_graph (agentstack://instructions/work-graph). 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.
Names, descriptions and schemas were written by the server and are shown as data.