





{"success":true,"content":"# Research, unlocks and useful capabilities\n\nTemplate revision: 7. This is the canonical research procedure. The [operations cycle](MCP_Agent_Operations.md) requires a brief review each session; consult this playbook when a decision, queue, job or unlock goal needs attention. Keep project facts in [current state](MCP_Agent_Current_State.md), outcomes in [goals](MCP_Agent_Goals.md), and evidence in [learning](MCP_Agent_Learning.md).\n\nDefault to manual design and comparison. The ordinary AI research assistant is optional for difficult problems when configured and useful. Preserve the user's class, personality, agreements and risk preferences. Research is a standing planning track, not a requirement to spend every session or unlock everything.\n\n## Define the capability\n\nStart with a need supported by observations: scouting, hauling, defense, ground delivery, an observed counter, cheaper sustainable production or an economic unlock. State the baseline, operational role, success measure, deadline or useful review event, acceptable loss and conditions that invalidate the proposal. Unknown enemy statistics remain unknown.\n\nInspect available, shared and queued designs before creating another. Compare using the current library, obtaining an eligible shared design, changing tactics, increasing TECH, improving facilities, researching alternatives and postponement. A useful existing design can be the correct research decision. Separate empire TECH, personal Components/tree nodes, weapon/unit blueprints and deployed forces. The [research libraries](Research_Libraries.md) distinguish available and pending work; nuclear research requires personal Nuclear Warfare access, not TECH alone.\n\n## Compare designs and complete costs\n\nFollow [specialized production and research worlds](MCP_Agent_Economy.md#specialized-production-and-research-worlds) when selecting sites. Quote weapon/unit research and queue payments at the laboratory hub, then quote the finished design's manufacture at the matching production hub. A ship can be researched on one world and built on another with more orbital facilities; nuclear research uses labs and nuclear manufacturing uses nuke plants. Retain separate research and production site IDs, local funding and delivery plans.\n\nFor each feasible candidate record the role, actual components, compatible weapons, intended quantity, research site, quotes, assumptions and expected operational benefit. Bound comparisons by the campaign's actual call/time budget; do not repeatedly optimize unchanged inputs without a decision.\n\n- **Weapons:** compare power with observed protection, damage with the target and loss objective, and accuracy, rate and area/range with the engagement. The designer's `area` field participates in range/engagement behavior as well as size. More of one attribute consumes fitting and production resources and does not guarantee a better weapon. Read the [combat/research basics](Basics_Of_Combat_And_Research.md), then use actual size, compatibility and pricing returned by quotes.\n- **Units:** compare hull and equipment capacity with mobility, protection, combat command, hangar, scanners, cloak and hyperunit needs. Do not confuse colony command centres with unit command. Positive raw engine is not proof of effective movement; see the [verified mechanics examples](MCP_Agent_Mechanics.md). Fit weapons by their actual IDs, allowed unit types and quantities.\n- **Whole operations:** a counter needs to arrive with suitable forces and support. A transport needs usable cargo, legal loading and a complete delivery route. A ground force needs delivery and reinforcement. A successful blueprint alone satisfies none of those tasks.\n\nUse weapon and unit research quotes and `query.production_quote`; include every distinct new weapon once, the unit research, production of the intended batch, required facilities, transfer fees, transport and replacement risk. Keep Credits and Exium separate, preserve accepted-operation and defense reserves, and quote changed sites/quantities again. Exact quotes are observations, not reservations.\n\nA manual unit quote takes real accessible weapon IDs, not hypothetical weapon objects. If a needed weapon does not exist, quote its size/cost/compatibility first and treat later unit prices as estimates until real IDs can be quoted. Fund a bounded weapon-research stage only when the remaining fitting and budget case is credible; verify the returned weapon before quoting/researching its dependent unit. Do not fabricate IDs or label an incompletely quoted package exact. The optional assistant can quote a combined unit and distinct-new-weapon package when available.\n\n**Hypothetical comparison, not game quotes:** an existing unit costs 900 Credits/90 Exium each. A candidate costs 600/80, plus 2,600/200 for a distinct new weapon and 3,900/300 for unit research. At 20 units the baseline costs 18,000/1,800, while the new package costs 18,500/2,100. Credit-only savings start at 22 units; Exium only breaks even at 50. At the planned quantity, research is not a cost saving in either currency, although a demonstrated new capability could justify it. This comparison assumes identical facilities/logistics and excludes any additional costs, which must be budgeted separately. Never invent an exchange rate to hide the tradeoff.\n\nBefore dispatch, record the chosen candidate, exact available quotes, remaining estimates, the current funded step, prediction and abort/replan conditions. Refresh after TECH, laboratory, ownership, weapon, rule or funding changes. Research and manufacture through ordinary MCP; verify actual deductions and returned or unambiguously resolved design IDs.\n\n## Research queues\n\nQueuing saves a pending design; it does not pay for or automatically complete its research. Track real remaining research points and current funding. Ordinary `advance_queued_weapon` and `advance_queued_unit` spend resources on an explicit positive number of research points at an owned site.\n\nComplete required weapons before the dependent unit. Inspect `query.weapon_inventory` and `query.unit_inventory` with the appropriate status and continuations, and use the normal design/research view to inspect pending work. A ready weapon is not proof the unit is complete; a queued unit cannot be counted as a manufactured force.\n\nRespect shared, read-only, deprecated and ownership restrictions. Before editing or removing pending work, inspect dependencies and investment already made; do not assume a refund or that paid progress survives a design change. Keep actual queue IDs, remaining points, responsible site, next payment and resolving event in current state. Do not wait for an imaginary background completion timer.\n\n## Tech-tree unlocks and activation\n\nMaintain the ranked [research and unlock backlog](MCP_Agent_Goals.md#research-and-unlock-backlog). Read `query.techtree` at an owned colony. Use the exact returned node key as `unlock_technology`'s `node_key`; inspect its current class branch, TECH threshold, Component cost, prerequisites and eligibility. Never substitute a remembered node name or assume receiving TECH transfers personal unlocks.\n\nPrioritize the useful complete benefit: a passive economic bonus, a needed research/sharing permission, logistics access or a specific military capability. Compare competing uses of scarce Components. Refresh after a battle, another unlock, class/empire change or shared TECH. Do not change the user's class just to fit a tree recipe.\n\nFor a useful funded eligible node, either execute `unlock_technology` with a journaled stable command ID or record a specific reason for deferral and its next review. Verify the receipt and the node's returned unlocked status. `canUnlock: false` alone is not proof of success; it can also mean a prerequisite is missing. Reconcile the actual personal Component balance and other intervening events. Already owned nodes are baseline capabilities; do not issue a new unlock merely to manufacture a receipt.\n\nComplete the **unlock goal** at the verified unlocked node. Keep **activation** as a separate dependent goal: fund/build needed structures, configure the ability, assign appropriate forces or produce/deploy the newly permitted design. Passive abilities need an applicable before/after observation; active abilities need verified configuration and later operational evidence when appropriate. Do not invent an enemy encounter merely to claim a test.\n\nFor example, a strategic reliable-delivery objective can contain separate goals to unlock Stargate Construction, build eligible useful gates, and verify an operational route. The first goal can be complete while the delivery capability is still blocked by construction funds or destination access.\n\n## Component expeditions\n\nComponents are personal salvage, not Credits, Exium, design fitting points or a shared alliance treasury. Reserve each available Component only once. Expected battle rewards are uncertain future income and cannot fund an immediate unlock.\n\nA Component acquisition goal is permitted when the unlock benefit justifies a bounded expedition within the campaign's mandate. Apply normal [combat planning](MCP_Agent_Combat.md), reconnaissance, protection, diplomacy, access, logistics, garrison and loss limits. Compare completing another justified objective, waiting, taking an economic prerequisite or choosing a different unlock. Do not promise a particular payout from a single target.\n\nThe [Components rules](Components.md) award salvage only after a decisive battle with a surviving winner. Enemy wreckage, participation shares, tech-gap variation, caps and counterattack adjustments affect the result. More own hull outside the fight is not more reward; a larger allied force can change the personal share. An ongoing battle or mutual destruction does not establish an award.\n\nRecord the target evidence, legal route, expected benefit, full expedition cost, maximum acceptable loss and fallback. After resolution inspect authorized reports and actual personal Components, distinguish a recorded payout from an estimate or unavailable old report, then revise the remaining gap and unlock priority. A successful expedition can still leave the unlock unfunded.\n\n## Optional AI-assisted research\n\nUse the configured player research assistant for a difficult design or report-informed counter when useful within the campaign's budget; it is optional, not a replacement for evaluating the result. It supports its advertised unit types. Manual research remains available for other legal designs.\n\nRecord the brief, requirements that must survive, authorized report/target, intended role, site, separate research/production budget caps, job ID and next status checkpoint. Generation, readiness, purchase, queue completion and deployment are different stages. Use the returned status; do useful independent work instead of endless polling or creating duplicate active jobs.\n\nReview the actual candidate, combined distinct-weapon/unit research cost, per-unit production estimate and current returned purchase quote. A proposed counter or internal qualification is not proof of victory against a hidden or different force. Quotes/jobs can become stale after TECH, empire/round, eligibility or rule changes; refresh or regenerate the saved brief as required. Reconcile existing job/command outcomes before retrying, cancelling or purchasing. Never repeat an uncertain purchase under a new ID.\n\nBuying immediately and adding to the ordinary queue have different consequences. Verify which happened, resulting IDs and actual charges. A queued AI result still needs its weapons and unit research finished through normal queue operations. Keep credentials and internal/provider diagnostics out of campaign records; player-visible findings are sufficient.\n\n## Validate and adopt\n\nPrefer a bounded initial batch when uncertainty warrants it and normal logistics permit it. State the predicted benefit, complete allocation, loss ceiling and rejection criterion. Verify the finished blueprint, actual production, deployment and later reports or delivery measurements. Manufacturing is not deployment; an unlock is not an activated capability.\n\nCompare the baseline under relevant conditions and comparable game-time intervals. Separate a design failure from wrong targets, poor composition, unsupported travel, supply interruptions or incomplete intelligence. Record costs, losses and revised explanation; improve, abandon or expand the design accordingly. Do not mass-produce a demonstrated ineffective design or deprecate useful existing designs before their replacement is available.\n\nAt handoff every active project has its real stage, IDs, remaining costs/points, owned funding site, protected reservations, evidence and next useful action or event. Keep rejected and uncertain intents distinct. Review useful ready unlocks and affordable queue steps each session while preserving urgent defense and actionable replies.\n\n## MCP operation reference\n\nInspect current schemas before using these operations; the table names operations, not complete argument sets. All commands require stable `command_id` values. Use ordinary player-visible intelligence and state only, never an administrative simulator or direct database access.\n\n| Tool family | Operation | Purpose |\n| --- | --- | --- |\n| `ge_research_read` | `query.unit_inventory` | Available and pending unit designs, with pagination. |\n| `ge_research_read` | `query.weapon_inventory` | Available and pending weapon designs. |\n| `ge_research_read` | `query.researchpanel` | Inspect the ordinary research panel and pending work. |\n| `ge_research_read` | `query.spaceunit` | Space design details; use the corresponding ground/soldier/nuke detail operation for that type. |\n| `ge_research_read` | `query.weapon` | Actual weapon details and compatibility. |\n| `ge_research_read` | `query.research_weapon_quote` | Weapon size, compatibility and research price. |\n| `ge_research_read` | `query.research_space_quote` | Space design validation, research and production estimates. |\n| `ge_research_read` | `query.research_ground_quote` | Ground design quote. |\n| `ge_research_read` | `query.research_soldier_quote` | Soldier design quote. |\n| `ge_research_read` | `query.research_nuke_quote` | Nuclear design quote under its unlock restrictions. |\n| `ge_military_read` | `query.production_quote` | Cost of the actual finished design and intended quantity. |\n| `ge_research_start` | `research_weapon` | Research or queue a legal weapon. |\n| `ge_research_start` | `research_space_unit` | Research or queue a legal space unit. |\n| `ge_research_start` | `research_ground_unit` | Research or queue a ground unit. |\n| `ge_research_start` | `research_soldier_unit` | Research or queue a soldier. |\n| `ge_research_start` | `research_nuke_unit` | Research or queue a nuclear design when eligible. |\n| `ge_research_queue` | `advance_queued_weapon` | Pay for a stated amount of remaining weapon research. |\n| `ge_research_queue` | `advance_queued_unit` | Pay for remaining unit research after required weapons. |\n| `ge_unit_production` | `produce_space_units` | Build an eligible finished design through normal production. |\n| `ge_unit_production` | `produce_ground_units` | Build eligible ground units. |\n| `ge_unit_production` | `produce_soldier_units` | Build eligible soldiers. |\n| `ge_player_read` | `query.operations_snapshot` | Observe actual owned forces, resources and coverage gaps. |\n| `ge_military_read` | `query.fleet` | Verify authorized space-force composition and orders. |\n| `ge_military_read` | `query.army` | Verify authorized ground-force composition and orders. |\n| `ge_research_read` | `query.techtree` | Personal branch, Components, live nodes and requirements. |\n| `ge_research_start` | `unlock_technology` | Unlock the exact eligible node at an owned colony. |\n| `ge_research_read` | `query.ai_research_status` | Inspect owned jobs, results and available purchase information. |\n| `ge_research_read` | `query.ai_research_targets` | Resolve permitted targets from an authorized report. |\n| `ge_ai_research` | `start_ai_research` | Start an optional bounded research job. |\n| `ge_ai_research` | `retry_ai_research` | Retry the appropriate owned job with a revised brief. |\n| `ge_ai_research` | `cancel_ai_research` | Cancel an owned job when no longer useful. |\n| `ge_ai_research` | `purchase_ai_research` | Purchase the current result or add it to the normal queue. |\n"}

