# Synchronex Full LLM Context

> Machine-readable reference for Synchronex docs, REST API, MCP tools, memory schema, and AI worker types.

Synchronex is an AI command center for solo founders and small teams. It stores persistent product memory, exposes that memory through REST and MCP, and runs role-based AI workers with explicit autonomy controls.

## API Endpoints

- GET /api/v1/memory/identity - Get Identity Kernel. Auth: bearer token with read scope.
- PUT /api/v1/memory/identity - Update Identity Kernel. Auth: bearer token with write scope.
- GET /api/v1/memory/product/{name}/state - Get Product State. Auth: bearer token with read scope.
- POST /api/v1/memory/product/{name}/state - Update Product State. Auth: bearer token with write scope.
- GET /api/v1/memory/search - Search memory and product state. Auth: bearer token with read scope.
- GET /api/v1/memory/export-mif - Export portable Memory Interchange Format. Auth: bearer token with read scope.
- POST /api/v1/memory/import-mif - Import portable Memory Interchange Format. Auth: bearer token with write scope.
- POST /api/v1/memory/consolidate - Run async memory consolidation. Auth: bearer token with write scope.
- GET /api/v1/memory/decisions/search - Search Decision Index. Auth: bearer token with read scope.
- POST /api/v1/memory/decisions - Create Decision. Auth: bearer token with write scope.
- DELETE /api/v1/memory/decisions/{id} - Delete Decision. Auth: bearer token with write scope.
- POST /api/v1/ingest/{project_id} - Ingest Project Signal. Auth: bearer token with write scope.
- POST /api/v1/signals/ingest - Ingest Signal Batch. Auth: bearer token with write scope.
- POST /api/v1/connectors/intake - Submit External AI Command. Auth: bearer token with write scope.
- GET /api/v1/ai/costs - Get AI Cost Summary. Auth: bearer token with read scope.
- GET /api/v1/ai/costs/{product} - Get Product AI Costs. Auth: bearer token with read scope.
- GET /api/v1/byok - List BYOK Credentials. Auth: bearer token with read scope.
- POST /api/v1/byok - Create BYOK Credential. Auth: bearer token with write scope.
- DELETE /api/v1/byok/{id} - Delete BYOK Credential. Auth: bearer token with write scope.
- GET /api/v1/kernel/export - Export Active Kernel. Auth: bearer token with read scope.
- POST /api/v1/kernel/export - Export Active Kernel. Auth: bearer token with read scope.
- GET /api/v1/scope - Get Active Scope. Auth: bearer token with read scope.
- PATCH /api/v1/scope - Set Active Scope. Auth: bearer token with write scope.
- POST /api/v1/ingest/events - Ingest Business Event. Auth: bearer token with write scope.
- POST /api/v1/ingest/datasets/{name}/rows - Ingest Dataset Rows. Auth: bearer token with write scope.
- POST /api/v1/ingest/upload - Preview Dataset File Upload. Auth: bearer token with write scope.
- POST /api/v1/ingest/upload/confirm - Confirm Dataset File Upload. Auth: bearer token with write scope.
- GET /api/v1/bridge/status - Get Agent Bridge Status. Auth: bearer token with read scope.
- POST /api/v1/bridge/generate - Queue or Run Agent Bridge Generation. Auth: bearer token with write scope.
- GET /api/v1/bridge/generate/{jobId} - Get Agent Bridge Generation Status. Auth: bearer token with read scope.
- POST /api/v1/render/video - Render Marketing Video. Auth: bearer token with write scope.
- POST /api/v1/render/producibles/jobs - Queue Producible Render Job. Auth: bearer token with write scope.
- GET /api/v1/render/producibles/jobs/{jobId}/status - Get Producible Render Job Status. Auth: bearer token with read scope.

## MCP Tools

- register_session - Register this AI session with Synchronex so it appears in the pane picker. Call this at the very start of every session. If you have a remote share URL (Claude Code: /share), pass it as session_url — Synchronex will offer it as a live pane in the terminal grid. Returns a session_id to use in subsequent calls.. Scope: write. Inputs: product_name, session_url, agent_name, label, metadata.
- end_session - Signal that this IDE session has ended. Synchronex will summarize decisions made during the session, update the Product State Document, and write a handoff note for the next session. Call this in your IDE teardown hook or at the end of each session. Returns decisions_captured and handoff_note for confirmation.. Scope: write. Inputs: session_id, summary, ended_by.
- get_identity_kernel - Returns the user's Identity Kernel — identity, product portfolio, standing constraints, and model preferences. Inject the returned injection_block at the start of every session to restore persistent identity context.. Scope: read. Inputs: session_id, client_type.
- get_product_state - Returns the current Product State Document for a specific product, including last session summary, handoff note, what works, blockers, and open questions.. Scope: read. Inputs: product_name.
- search_decisions - Semantic search over the cross-product decision history. Returns ranked, relevant decisions trimmed to a token budget. Falls back to keyword search if embeddings are unavailable.. Scope: read. Inputs: query, product_name, limit, token_budget.
- get_session_handoff - Returns the handoff note (next_session_starts_here) for a product. Use this at the very start of a session to orient before loading full state. The lightest memory call available.. Scope: read. Inputs: product_name.
- check_memory_confidence - Audits a session context block for stale data, constraint violations, and overall confidence score. Paste your CLAUDE.md or any injected memory block to validate it before relying on it.. Scope: read. Inputs: context.
- log_decision - Writes a new architectural or product decision to the cross-product decision index. Auto-tags, generates an embedding, and calculates temporal weight. Use this when a significant decision is made during a session.. Scope: write. Inputs: decision, reasoning, product_name, affects_products, workspace_id.
- log_project_event - Record a project signal or event into the Synchronex command center. Use for deployments, errors, decisions, and any notable state change.. Scope: write. Inputs: project_id, event_type, title, severity, body, occurred_at.
- send_message - Send a team message or direct message to a teammate. Use this when the user explicitly asks you to message someone on their team.. Scope: write. Inputs: to, body, entity_reference, workspace_id.
- export_kernel - Exports the authenticated user's active kernel into client-specific formats or a registered scaffold template.. Scope: read. Inputs: format, template_id, product_name, budget, session_id, client_type.
- memory_kernel_attest - Confirms that the AI agent loaded the Synchronex Identity Kernel for this session. Call once after reading the exported kernel or receiving kernel context.. Scope: read. Inputs: session_id, sentinel_token, kernel_fingerprint.
- update_product_state - Updates the Product State Document for a specific product. Accepts partial updates — only provided fields are changed. Creates the state document if it doesn't exist yet.. Scope: write. Inputs: product_name, last_session_summary, next_session_starts_here, status, what_works, what_is_blocked, open_questions.
- studio_generate - Queue brand-aware content generation. Video requests use the async studio job pattern; poll with studio_get_job.. Scope: write. Inputs: content_type, brief, format, workspace_id, brand_id, provider, metadata.
- studio_plan - Queue a brand-aware content lane plan from a brief.. Scope: write. Inputs: brief, duration_weeks, workspace_id, brand_id.
- studio_propose - Queue content variants for approval from an existing generation request.. Scope: write. Inputs: generation_request_id, variant_count.
- studio_get_job - Poll a studio generation job or async media job.. Scope: read. Inputs: job_id.
- worker_list - List active Synchronex workers and recent run state.. Scope: read. Inputs: status, workspace_id.
- worker_run - Trigger an active worker run. Returns immediately with a run id; poll with worker_get_run.. Scope: write. Inputs: worker_type, task, workspace_id, brand_id, input.
- worker_get_run - Read the status and output of a worker run.. Scope: read. Inputs: run_id.
- worker_approve - Approve an escalated worker run and resume execution. Requires full-scope MCP credentials.. Scope: full. Inputs: run_id, decision_reason, edited_action.
- worker_set_schedule - Set, change, or disable a worker's run schedule. Accepts a cadence preset (daily/weekly/biweekly/monthly) or an arbitrary 5-field cron expression (minimum interval: hourly). Resolves the worker by worker_id, or by worker_type + workspace_id when unambiguous. Returns the resulting schedule with next_fire_at so the companion can confirm it to the user in plain language.. Scope: write. Inputs: worker_id, worker_type, workspace_id, cadence, cron, timezone, disable.
- worker_get_schedule - Read a worker's current run schedule (cadence/cron, timezone, next_fire_at, active state). Resolves the worker by worker_id, or by worker_type + workspace_id when unambiguous.. Scope: read. Inputs: worker_id, worker_type, workspace_id.
- kernel_get - Read the authenticated user Identity Kernel.. Scope: read. Inputs: none.
- kernel_update_slice - Merge a validated partial update into one Identity Kernel slice.. Scope: write. Inputs: slice_name, updates.
- kernel_list_products - List products managed by the authenticated user.. Scope: read. Inputs: none.
- ceo_brief - List pending CEO proposals and escalated worker runs awaiting review.. Scope: read. Inputs: limit.
- ceo_approve - Approve a CEO proposal or escalated worker run. Requires full-scope MCP credentials.. Scope: full. Inputs: proposal_id, run_id, notes.
- ceo_reject - Reject a CEO proposal or escalated worker run with feedback. Requires full-scope MCP credentials.. Scope: full. Inputs: proposal_id, run_id, feedback.
- list_datasets - Lists normalized business datasets available to the authenticated user.. Scope: read. Inputs: workspace_id.
- query_dataset - Queries rows from a normalized business dataset by exact-match filters.. Scope: read. Inputs: name, limit, filters, workspace_id.
- codebase_read_file - Read a UTF-8 file from a GitHub-backed registered project. Use before proposing code edits.. Scope: read. Inputs: project_id, path, ref.
- codebase_write_file - Create or update one UTF-8 file in a GitHub-backed registered project. Requires full MCP scope.. Scope: full. Inputs: project_id, path, content, branch, message, sha.
- codebase_apply_diff - Apply a prepared file-set diff to a GitHub-backed registered project by writing complete file contents. Requires full MCP scope.. Scope: full. Inputs: project_id, branch, message, files.
- codebase_open_pull_request - Open a GitHub pull request for a registered project and track it in Synchronex. Requires full MCP scope.. Scope: full. Inputs: project_id, title, head, base, body, draft, product_id, worker_run_id, workstream_brief_id.
- codebase_list_pull_requests - List Synchronex-tracked pull requests for a GitHub-backed registered project.. Scope: read. Inputs: project_id, state, limit.
- memory_kernel_read - Returns the authenticated user Identity Kernel as structured JSON.. Scope: read. Inputs: none.
- memory_kernel_update - Merges a partial update into the authenticated user Identity Kernel.. Scope: write. Inputs: updates.
- memory_search - Searches memory layers (kernel, product state, decisions, free-form memories); decisions and free-form memories use semantic pgvector search.. Scope: read. Inputs: query, limit, layer.
- memory_product_state_read - Returns the current state document for a product by name.. Scope: read. Inputs: product.
- memory_product_state_write - Applies partial updates to a product state document.. Scope: write. Inputs: product, updates.
- memory_decision_log - Appends a cross-product decision to the decision index.. Scope: write. Inputs: decision, context, products_affected.
- memory_session_summary - Stores a cross-AI session summary with attribution.. Scope: write. Inputs: summary, topics, ai_tool.
- remember_free_form - Stores a free-form note in cloud memory with optional tags and workspace scope.. Scope: write. Inputs: title, content, tags, workspace_id, is_pinned.
- recall_free_form - Searches free-form cloud memories by semantic or keyword match.. Scope: read. Inputs: query, limit, workspace_id.
- list_registered_workspaces - Lists registered Synchronex workspace sync targets for the authenticated user.. Scope: read. Inputs: none.
- get_drift_events - Lists workspace drift events for the authenticated user.. Scope: read. Inputs: status.
- approve_inbound_change - Accepts or rejects a pending workspace drift event.. Scope: write. Inputs: action, note, resolved_by, drift_event_id.
- sync_workspace_state - Renders the current kernel export for a registered workspace. The MCP client writes the returned file content locally.. Scope: write. Inputs: workspace_registration_id.
- intake_command - Submit an inbound command through the intake spine. The command is recorded to the immutable command ledger and queued as a PENDING proposal for human approval; nothing executes directly.. Scope: write. Inputs: raw_content, idempotency_key, requested_capability.
- get_inbound_commands - Lists inbound command proposals for the authenticated user.. Scope: read. Inputs: status.
- approve_inbound_command - Accepts or rejects a pending inbound command proposal. Approval is a recorded decision only; execution is separately gated. Requires full-scope MCP credentials.. Scope: full. Inputs: proposal_id, action, note.
- chief_of_staff_brief - Answer what needs attention using CEO/systems-orchestrator context, worker recommendations, approvals, and jurisdiction routing.. Scope: read. Inputs: workspace_id, question.
- read_attention_feed - Read the companion-worker attention feed: today, approvals, worker feed, and watching lanes.. Scope: read. Inputs: workspace_id, limit.
- read_opportunities - Read ranked companion-worker opportunities from worker_signal_ledger.. Scope: read. Inputs: workspace_id, limit.
- create_opportunity - Create a normalized companion-worker opportunity in worker_signal_ledger.. Scope: write. Inputs: workspace_id, worker_type, signal_type, title, summary, why_now, confidence, severity, entity_type, entity_id, recommended_action, evidence.
- draft_worker_action - Draft a worker action for user approval. Alias of submit_proposed_action_for_approval.. Scope: write. Inputs: workspace_id, worker_id, worker_run_id, worker_type, source_signal_id, opportunity_key, action.
- submit_proposed_action_for_approval - Submit a companion-worker proposed action into the existing approval-first deliverable flow.. Scope: write. Inputs: workspace_id, worker_id, worker_run_id, worker_type, source_signal_id, opportunity_key, action.
- read_approval_status - Read pending approval status for worker outputs and CEO briefings.. Scope: read. Inputs: workspace_id, output_id, briefing_id, limit.
- read_worker_outputs - Read approval-bound worker outputs for a connected agent.. Scope: read. Inputs: workspace_id, output_id, briefing_id, limit.
- read_watch_config - Read companion-watch configuration stored on workers.config.companion_watch.. Scope: read. Inputs: workspace_id, worker_type.
- update_watch_config - Update companion-watch configuration for an existing worker.. Scope: write. Inputs: workspace_id, worker_id, companion_watch.
- subscribe_to_feed - Attach a public HTTPS feed to a worker for periodic ingestion and attention recommendations.. Scope: write. Inputs: workspace_id, worker_id, feed_url, feed_type, display_label, description, fetch_interval_minutes.
- unsubscribe_from_feed - Remove a worker feed subscription and its ingested items.. Scope: write. Inputs: workspace_id, source_id.
- read_feed_sources - List external feed subscriptions and their latest fetch state.. Scope: read. Inputs: workspace_id, worker_id, include_disabled.
- read_feed_items - Read normalized external feed items available to workers and the attention system.. Scope: read. Inputs: workspace_id, worker_id, source_id, limit.
- trigger_feed_fetch - Fetch one owned feed source now, deduplicate its items, and emit worker attention opportunities.. Scope: write. Inputs: workspace_id, source_id.
- gmail_search - Search Gmail threads in one explicitly labeled connected account.. Scope: read. Inputs: account_label, q, limit, workspace_id.
- gmail_read - Read a full Gmail thread from one explicitly labeled connected account.. Scope: read. Inputs: account_label, thread_id, workspace_id.
- gmail_send - Propose a new Gmail message from one explicitly labeled account. The message is not sent until a human approves the connector write decision.. Scope: write. Inputs: account_label, to, subject, body, workspace_id.

## Identity Kernel Schema

- v: schema version, currently 1
- identity: name, company, location, timezone, role, industry
- products: up to 10 product snapshots with name, optional description, status, and primary flag
- constraints: address_as, communication_style, always, never, coined_terms
- model_preferences: available providers and per-task routing preferences
- injection_text: pre-rendered memory block, capped at 500 tokens for fast session injection

## Memory Layers

- Identity Kernel: stable user and business profile injected into every session.
- Product State: per-product status, blockers, open questions, session summary, and next-session handoff.
- Decision Index: timestamped decisions with reasoning, tags, affected products, semantic search, and temporal weighting.

## AI Workers

- marketing (Marketing Strategist) - Plans campaigns, writes copy, manages email outreach, and grows your audience via Sendinel.. Autonomy: supervised. Trigger: scheduled.
- admin_crm (Admin & CRM Manager) - Manages contacts, tracks deals, updates CRM records, and handles follow-up communications.. Autonomy: supervised. Trigger: scheduled.
- cfo (CFO) - Tracks revenue, monitors burn rate, analyzes Stripe data, and generates financial summaries.. Autonomy: standard. Trigger: scheduled.
- ceo (CEO) - Cross-product strategic overview, decision support, and executive intelligence synthesis.. Autonomy: standard. Trigger: scheduled.
- devsecops (DevSecOps Engineer) - Infrastructure management, security auditing, deployment monitoring, and incident response.. Autonomy: supervised. Trigger: scheduled.
- product_manager (Product Manager) - Feature prioritization, roadmap, and issue tracking.. Autonomy: supervised. Trigger: on_demand.
- frontend_engineer (Frontend Engineer) - UI implementation and code review.. Autonomy: supervised. Trigger: on_demand.
- seo_content (SEO & Content) - Organic growth, content creation, and keyword research.. Autonomy: supervised. Trigger: on_demand.
- systems_orchestrator (OrchestrAItor) - Coordinates complex multi-step work by delegating to specialist workers.. Autonomy: standard. Trigger: on_demand.
- customer_success (Customer Success) - User retention, health scoring, and churn prevention.. Autonomy: supervised. Trigger: on_demand.
- pricing_strategist (Pricing Strategist) - Monetization design, tier structure, and pricing optimization.. Autonomy: supervised. Trigger: on_demand.
- sales_consultant (Sales Consultant) - Prospect qualification, outreach, and deal closing.. Autonomy: supervised. Trigger: on_demand.
- content_production (Content Production) - Content calendar, brand consistency, and multi-channel publishing.. Autonomy: supervised. Trigger: on_demand.
- distribution_manager (Distribution Manager) - Content and product distribution across newsletters, social, and partnerships.. Autonomy: supervised. Trigger: on_demand.
- sendinel_producer (Sendinel Producer) - Analyzes Sendinel campaign, deliverability, segment, and engagement signals for CEO-gated recommendations.. Autonomy: supervised. Trigger: scheduled.
- competitor_analyst (Competitive Intelligence Analyst) - Proactively compares your product against known competitors and surfaces gaps, advantages, and strategic risks.. Autonomy: standard. Trigger: scheduled.
- hr_manager (People Ops Manager) - Tracks hiring, onboarding, payroll, benefits, PTO, and contractor admin for a growing team.. Autonomy: supervised. Trigger: scheduled.
- legal_ops (Legal Ops Manager) - Tracks contracts, NDAs, vendor agreements, signature status, renewals, and legal admin.. Autonomy: supervised. Trigger: scheduled.
- data_analyst (Data Analyst) - Monitors product usage, funnels, cohorts, attribution, and business metric anomalies.. Autonomy: standard. Trigger: scheduled.
- operations_manager (Operations Manager) - Coordinates projects, deadlines, blockers, team workload, and cross-functional follow-through.. Autonomy: standard. Trigger: scheduled.
- seo_strategist (SEO Strategist) - Crawls your product site, scores technical SEO health, builds an ICP-aligned keyword strategy, and generates AEO-ready content briefs.. Autonomy: supervised. Trigger: scheduled.

## Operating Contracts

- supervised: drafts or recommendations require explicit user approval before external action.
- standard: lower-risk work can continue through classifier-approved paths; risky handoffs require approval.
- autonomous: approved background work logs actions and decisions, with notifications on failure or policy boundaries.

## Key Docs

- https://synchronex.ai/docs/quickstart
- https://synchronex.ai/docs/authentication
- https://synchronex.ai/docs/api
- https://synchronex.ai/docs/mcp
- https://synchronex.ai/docs/mcp/tools
- https://synchronex.ai/docs/cloud-developer-mode
- https://synchronex.ai/docs/concepts/memory
- https://synchronex.ai/docs/concepts/workers